CN107301383A - A kind of pavement marking recognition methods based on Fast R CNN - Google Patents

A kind of pavement marking recognition methods based on Fast R CNN Download PDF

Info

Publication number
CN107301383A
CN107301383A CN201710421849.XA CN201710421849A CN107301383A CN 107301383 A CN107301383 A CN 107301383A CN 201710421849 A CN201710421849 A CN 201710421849A CN 107301383 A CN107301383 A CN 107301383A
Authority
CN
China
Prior art keywords
pavement marking
recognition methods
cnn
fast
layer
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN201710421849.XA
Other languages
Chinese (zh)
Other versions
CN107301383B (en
Inventor
刘兰馨
李巍华
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
South China University of Technology SCUT
Original Assignee
South China University of Technology SCUT
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by South China University of Technology SCUT filed Critical South China University of Technology SCUT
Priority to CN201710421849.XA priority Critical patent/CN107301383B/en
Publication of CN107301383A publication Critical patent/CN107301383A/en
Application granted granted Critical
Publication of CN107301383B publication Critical patent/CN107301383B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

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
    • G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
    • G06V20/582—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads of traffic signs
    • 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
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/07—Target detection

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Multimedia (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Engineering & Computer Science (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a kind of pavement marking recognition methods based on Fast R CNN, including step：IMAQ and pretreatment are carried out, sample set is made；Training set is inputted, multitask training Fast R CNN networks；Picture to be identified obtains characteristic pattern by some convolutional layers and pond layer；Corresponding feature frame is obtained by candidate frame, via ROI ponds layer and full articulamentum, classification score is respectively obtained and window returns two output vectors；All results are handled by non-maxima suppression and produce final Target detection and identification result, traffic sign is recognized.The present invention has evaded the feature extraction operation of redundancy in the convolutional neural networks R CNN of region using this deep learning method of Fast R CNN, realizes multitask training, it is not required that extra characteristic storage space, improves detection speed and precision.Compared to shallow-layer Study strategies and methods, it has higher learning efficiency and accuracy of identification.

Description

A kind of pavement marking recognition methods based on Fast R-CNN
Technical field
The invention belongs to image procossing and automotive safety auxiliary driving field, Fast R-CNN are based on more particularly, to one kind Pavement marking detection and recognition methods, asked to solve the problems, such as in pavement marking identification that accuracy of identification is not high Topic.
Background technology
Traffic sign recognition (TSR, Traffic Signs Recognition) is used as one in vehicle-mounted accessory system Important branch, is one of current still unsolved problem.Due to containing many important transport information in traffic sign, such as to working as The speed prompt of preceding driving, the change of road ahead situation, driving behavior restriction, therefore in the accessory system, how soon Speed, accurately and efficiently identify the traffic sign in road and it fed back into human pilot or control system, for guarantee drive Safety is sailed, avoiding traffic accident has highly important Research Significance.
The conventional method of pavement marking identification includes the recognition methods based on shape, and feature extraction and classifying device is combined Method, the recognition methods of deep learning.Recognition methods robustness based on shape is poor, and effect is not good in complex environment.It is special Levy and extract the method recognition effect combined with grader preferably, but computing cost is big, and adaptive capacity to environment is poor.Deep learning Directly original image can be identified, the recessive character of reflection data essence be extracted, with enough study depth.Convolution Neutral net has the shared characteristic of local weight, and certain for all having situations such as environment is complicated, multi-angle changes is real-time Property and robustness.Accordingly, it would be desirable to which designing a kind of can accurately obtain the recognition methods of pavement marking in road scene.Ross B.Girshick proposed Fast R-CNN algorithms in 2015, had evaded R-CNN (Region-based Convolutional Neural Network) in redundancy feature extraction operation, realize multitask training, it is not required that extra characteristic storage is empty Between, improve detection speed and precision.
The content of the invention
To solve the above mentioned problem that prior art is present, the present invention will design a kind of road traffic based on Fast R-CNN Sign, can accurately obtain pavement marking in road scene, to contribute to auxiliary driver in complexity Under conditions of preferably perceive car external environment, prevent the generation of traffic accident.
To achieve these goals, technical scheme is as follows：
A kind of pavement marking recognition methods based on Fast R-CNN, including step：
IMAQ and pretreatment are carried out, sample set is made；
Training set is inputted, multitask training Fast R-CNN networks；
Picture to be identified is inputted into Fast R-CNN networks, by some convolutional layers and pond layer, characteristic pattern is obtained；
Several candidate frames are extracted using selection search (Selective Search) algorithm, arrived according to candidate frame in artwork Characteristic pattern mapping relations, find the corresponding feature frame of each candidate frame in characteristic pattern, and by each feature in the layer of ROI ponds Frame pondization arrives fixed size；
Feature frame is fixed to the characteristic vector of size by full articulamentum, the characteristic vector connects entirely via respective Layer is connect, classification score is respectively obtained and window returns two output vectors；
All results are subjected to non-maxima suppression processing and produce final Target detection and identification result, road traffic mark Will is recognized.
Further, the step of described image is gathered specifically includes：
Open vehicle-mounted traveling recorder, captured in real-time road traffic video information；
The video information that video camera is photographed carries out sub-frame processing, obtains an image collection sequence；
Image collection is screened, the image for including pavement marking is chosen.
Further, the step of described image pretreatment and sample set make specifically includes：
In the image of selection, target area is taken out, and zooms to 224 × 224 fixed sizes, for enhancing contrast Degree, then contrast enhancement processing is passed through into target area, original training set is obtained, test set is handled in the same way；
By original training set by rotation [- 12 °, 12 °], scale after [0.4,1.6], add initial data and concentrate, composition New training set；
It is random in new data set to take out the sample composition checking collection suitable with test set number, remaining sample composition Final training set.
Further, the Fast R-CNN network structures include：13 convolutional layers, 4 pond layers, 1 ROI pond Layer, 2 full articulamentums and two sane level layers.
Further, by the fixed size in each feature frame pond to 7 × 7 in the layer of the ROI ponds.
Further, the full connection output of described multitask training Fast R-CNN networks includes Liang Ge branches：cls_ Score layers and bbox_pred layers, described cls_score layers is used to classify, and described bbox_pred layers is used to adjust candidate frame position Put.
Further, when the characteristic vector is via respective full articulamentum, during which by singular value decomposition (SVD, Singular Value Decomposition) accelerate, respectively obtain two output vectors, i.e. the classification score of Softmax and Bounding-box windows are returned.
Further, to the Liang Ge branches of full connection output, the classification layer of output layer is trained using stochastic gradient descent method With recurrence layer until classification and the loss function convergence returned.
Further, the step of progress non-maxima suppression processing by all results is specifically included：According to output Each type objects are carried out non-maxima suppression respectively using window score and reject overlapping candidate frame, finally given by Liang Ge branches The window of revised highest scoring is returned in each classification.
Further, the pavement marking includes straight trip arrow, the to the left arrow that turns around, arrow, right-hand arrow, straight trip To the left arrow, straight trip right-hand arrow, rhombus graticule.
Compared with prior art, the present invention is in order to solve above-mentioned at least some problems of the prior art, it is proposed that Yi Zhongji Road traffic sign detection and recognition methods in Fast R-CNN.This method self manufacture pavement marking data set, passes through depth Degree study can extract the recessive character of reflection data essence from sample learning feature, with higher learning efficiency and Accuracy of identification, improves the robustness of detection algorithm, effectively increases the accuracy of pavement marking detection.
Brief description of the drawings
The invention provides accompanying drawing further understanding in order to disclosure, accompanying drawing constitutes the part of the application, But it is only used for illustrating the non-limiting example for some inventions for embodying inventive concept, rather than for making any limit System.
Fig. 1 is the pavement marking recognition methods based on Fast R-CNN of some example embodiments according to the present invention Flow chart.
Fig. 2 is the Fast R-CNN network structures of some example embodiments according to the present invention.
Fig. 3 is to train cost function schematic diagram according to the multitask of some example embodiments of the present invention.
Fig. 4 is the schematic diagram of the part traffic sign sample set of some example embodiments according to the present invention.
Fig. 5 is the pavement marking recognition methods based on Fast R-CNN of some example embodiments according to the present invention Testing result schematic diagram.
Embodiment
The present invention is elaborated with technical scheme below in conjunction with the accompanying drawings.
It is the flow chart of the pavement marking recognition methods based on Fast R-CNN, tool of the invention as shown in Figure 1 Body embodiment is：
A kind of pavement marking recognition methods based on Fast R-CNN, including step：
IMAQ and pretreatment are carried out, sample set is made；
Training set is inputted, multitask training Fast R-CNN networks；
Picture to be identified is inputted into Fast R-CNN networks, by some convolutional layers and pond layer, characteristic pattern is obtained；
About 2000 candidate frames are extracted using selection search (Selective Search) algorithm, according to candidate frame in artwork To characteristic pattern mapping relations, find the corresponding feature frame of each candidate frame in characteristic pattern, and ROI (area-of-interest, Region of Interest) each feature frame pondization arrived into fixed size in the layer of pond；
The feature frame is fixed to the characteristic vector of size by full articulamentum, via respective full articulamentum, point Classification score is not obtained and window returns two output vectors；
All results are handled by non-maxima suppression and produce final Target detection and identification result, traffic sign is obtained To recognize.
In certain embodiments, the step of described image is gathered specifically includes：
Open vehicle-mounted traveling recorder, captured in real-time road traffic video information, the video shot from drive recorder Resolution ratio is 1280*720 video image；
Sub-frame processing is carried out to captured video image, an image collection sequence is obtained；
Image collection is screened, 7 kinds of more pavement markings of occurrence number are chosen therefrom.
Specifically, the step of described image pretreatment and sample set make specifically includes：
In the image of selection, target area is taken out, and zooms to 224 × 224 fixed sizes, for enhancing contrast Degree, then contrast enhancement processing is passed through into target area, original training set is constituted, test set is handled in the same way；
By original training set by rotation [- 12 °, 12 °], scale after [0.4,1.6], add initial data and concentrate composition new Training set；
It is random in new data set to take out the sample composition checking collection suitable with test set number, remaining sample composition Final training set.
In the example depicted in fig. 4, selected traffic sign can be divided into 7 classes, respectively keep straight on arrow, turn around arrow Head, to the left arrow, right-hand arrow, keep straight on arrow, straight trip right-hand arrow, rhombus graticule to the left, number be expressed as 01,02 respectively, 03,04,05,06,07, recognition result is exported in this way.
K.Simonyan et al. is in document " K.Simonyan and A.Zisserman.Very deep Convolutional networks for large-scale image recognition, 2015. " the middle VGG16 proposed Network, including 13 convolutional layers, 5 pond layers and 3 full articulamentums.On the basis of VGG16, replaced with ROI ponds layer Last layer of pond layer of VGG-16 networks, two parallel layer replace above-mentioned VGG-16 networks the full articulamentum of last layer and Softmax layers.
As shown in Fig. 2 it is as follows to finally give Fast R-CNN network structures：Including 13 convolutional layers, 4 pond layers, 1 ROI ponds layer, 2 full articulamentums and two sane level layers.Size inputs network for 224 × 224 image pattern through input layer；It is right In all convolutional layers, convolution kernel size is 3 × 3, and step-length is 1；For all pond layers, 2 × 2 sampling sizes, Chi Hua are used Mode uses maximum pond.Activation primitive is using the linear unit activating (ReLU, Rectified Linear Units) of amendment Function, possesses the sparse ability of guiding appropriateness, can accelerate the training speed of network, precision is improved, it is to avoid what gradient disappeared Problem.
Original layer parameter will be initialized by pre-training mode.Full articulamentum for classification is using average as 0, standard deviation Initialized for 0.01 Gaussian Profile；Full articulamentum for recurrence is using average as 0, and standard deviation is at the beginning of 0.001 Gaussian Profile Beginningization, biasing is initialized to 0.
When tuning is trained, N full pictures are firstly added, the R candidate frame chosen from N pictures is then added. R/N candidate frame convolution of same image is shared to be calculated and internal memory, reduces computing overhead.The composition of R candidate frame is as follows：With Some true value overlap [0.5,1] candidate frame we be defined as prospect, account for the 25% of total amount；The maximum overlapping with true value exists [0.1,0.5) candidate frame be defined as background, account for the 75% of total amount.
ROI ponds layer averagely splits each feature frame according to fixed dimension, and maximum pond is carried out to every piece, can be by feature Feature frame not of uniform size is changed into the unified data of size on figure, sends into next layer.
Fast R-CNN networks carry out multitask training, and Classification Loss and recurrence loss are as shown in Figure 3.Cls_score layers For classifying, K+1 dimension group p are exported, expression belongs to K type objects and the probability of background, herein according to the number of detection classification, if Determine K=7；Bbox_pred layers are used to adjust candidate region location, 4*K dimension groups are exported, when expression is belonging respectively to K classes, it should flat The parameter of scaling is moved, for each classification one can be trained individually to return device.
Loss_cls layers are assessed classification cost Lcls, by the corresponding Probability ps of the u that truly classifiesuDetermine：
Lcls=-log pu (1)
Loss_bbox, which is assessed, returns loss cost Lloc, relatively truer classification u is corresponding to predict translation zooming parameterWith true translation zooming parameter v=(vx,vy,vw,vh) gap：
Combining classification loses and returned loss, and the network fine setting stage, total loss function was：
Agreement u=0 is background class, and background candidate region is that negative sample is not involved in returning loss, it is not necessary to candidate regions Domain carries out recurrence operation；λ controls loss and the balance for returning loss, λ=1.
Stochastic gradient descent method training network is utilized according to the loss function, until L convergences.
Being decomposed in Fast R-CNN networks using SVD accelerates full articulamentum to calculate；Object classification and window are returned Realized by full articulamentum, if full articulamentum input data is x, output data is y, full connection layer parameter is W, a forward direction Propagation is：
Y=Wx (5)
W is subjected to SVD decomposition, then propagated forward originally resolves into two steps：
Y=Wx=U (∑ VT) x=Uz (6)
U and V are intermediate variable, and this decomposition can greatly reduce amount of calculation, so as to realize full articulamentum speed-up computation.
Fig. 5 is the testing result schematic diagram of pavement marking, it is seen that effect is detected and recognized under general pavement conditions Fruit is good.
To sum up, the present invention proposes a kind of road traffic sign detection based on Fast R-CNN and recognition methods.This method is led to Deep learning is crossed from sample learning feature, the recessive character of reflection data essence can be extracted, imitated with higher study Rate and accuracy of identification, improve the robustness of detection algorithm, effectively increase the accuracy of pavement marking detection.Can be very The detection that the factor such as big degree alleviates pavement marking serious shielding, serious wear, deformation is serious, illumination variation is serious is brought Difficult the problem of.Part Methods step herein and flow may need to be performed by computer, so as to hardware, software, consolidate Part and its any combination of mode are implemented.
The preferred embodiments of the present invention are above are only, implementation and the interest field of invention is not intended to limit, it is all according to this Equivalence changes, modification, replacement that content described in patent application scope of patent protection is made etc., all should be included in the present patent application In the scope of the claims.It would be recognized by those skilled in the art that without departing from the scope and spirit of the present invention, can be wider It is changed and modified in wealthy each side.

Claims (10)

1. a kind of pavement marking recognition methods based on Fast R-CNN, it is characterised in that including step：
IMAQ and pretreatment are carried out, sample set is made；
Training set is inputted, multitask training Fast R-CNN networks；
Picture to be identified is inputted into Fast R-CNN networks, by some convolutional layers and pond layer, characteristic pattern is obtained；
Several candidate frames are extracted using selection searching algorithm, according to candidate frame in artwork to characteristic pattern mapping relations, in feature The corresponding feature frame of each candidate frame is found in figure, and each feature frame pondization is arrived into fixed size in the layer of ROI ponds；
The feature frame is fixed to the characteristic vector of size by full articulamentum, the characteristic vector connects entirely via respective Layer is connect, classification score is respectively obtained and window returns two output vectors；
All results are subjected to non-maxima suppression processing and produce final Target detection and identification result, pavement marking is obtained To recognize.
2. pavement marking recognition methods as claimed in claim 1, it is characterised in that：The step of described image is gathered is specific Including：
Open vehicle-mounted traveling recorder, captured in real-time road traffic video information；
The video information that video camera is photographed carries out sub-frame processing, obtains an image collection sequence；
Image collection is screened, the image for including pavement marking is chosen.
3. pavement marking recognition methods as claimed in claim 1, it is characterised in that：Described image is pre-processed and sample set The step of making, specifically includes：
In the image of selection, target area is taken out, and zooms to 224 × 224 fixed sizes, to strengthen contrast, then Contrast enhancement processing is passed through into target area, original training set is obtained, test set is handled in the same way；
Original training set through and rotation [- 12 °, 12 °], after scaling [0.4,1.6], initial data is added and concentrates, constitute newly Training set；
Random in new data set to take out the sample composition checking collection suitable with test set number, remaining sample composition is final Training set.
4. pavement marking recognition methods as claimed in claim 1, it is characterised in that the Fast R-CNN network structures Including：13 convolutional layers, 4 pond layers, 1 ROI ponds layer, 2 full articulamentums and two sane level layers.
5. pavement marking recognition methods as claimed in claim 1, it is characterised in that：Will be each in the layer of the ROI ponds Fixed size of the feature frame pond to 7 × 7.
6. pavement marking recognition methods as claimed in claim 1, it is characterised in that：Described multitask training Fast The full connection output of R-CNN networks includes Liang Ge branches：Cls_score layers and bbox_pred layers, the cls_score layers of use In classification, described bbox_pred layers is used to adjust candidate frame position.
7. pavement marking recognition methods as claimed in claim 1, it is characterised in that：The characteristic vector is via respective During full articulamentum, during which accelerate by singular value decomposition (SVD, Singular Value Decomposition), respectively obtain Two output vectors, i.e. the classification score and Bounding-box windows of Softmax are returned.
8. pavement marking recognition methods as claimed in claim 6, it is characterised in that：To two points of full connection output Branch, trains the classification layer of output layer using stochastic gradient descent method and returns layer until classification and the loss function returned are restrained.
9. pavement marking recognition methods as claimed in claim 6, it is characterised in that described that all results are subjected to non-pole The step of big value suppresses processing specifically includes：According to the Liang Ge branches of output, each type objects are entered respectively using window score Row non-maxima suppression rejects overlapping candidate frame, finally gives the window that revised highest scoring is returned in each classification.
10. pavement marking recognition methods as claimed in any one of claims 1-9 wherein, it is characterised in that hand on the road surface Logical mark includes straight trip arrow, the to the left arrow that turns around, arrow, right-hand arrow, keep straight on arrow, straight trip right-hand arrow, rhombus to the left Graticule.
CN201710421849.XA 2017-06-07 2017-06-07 Road traffic sign identification method based on Fast R-CNN Active CN107301383B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710421849.XA CN107301383B (en) 2017-06-07 2017-06-07 Road traffic sign identification method based on Fast R-CNN

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710421849.XA CN107301383B (en) 2017-06-07 2017-06-07 Road traffic sign identification method based on Fast R-CNN

Publications (2)

Publication Number Publication Date
CN107301383A true CN107301383A (en) 2017-10-27
CN107301383B CN107301383B (en) 2020-11-24

Family

ID=60134793

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710421849.XA Active CN107301383B (en) 2017-06-07 2017-06-07 Road traffic sign identification method based on Fast R-CNN

Country Status (1)

Country Link
CN (1) CN107301383B (en)

Cited By (38)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107895191A (en) * 2017-10-30 2018-04-10 上海寒武纪信息科技有限公司 A kind of information processing method and Related product
CN107909044A (en) * 2017-11-22 2018-04-13 天津大学 A kind of demographic method of combination convolutional neural networks and trajectory predictions
CN108009581A (en) * 2017-11-30 2018-05-08 中国地质大学（武汉） A kind of method for crack based on CNN, equipment and storage device
CN108095746A (en) * 2017-12-21 2018-06-01 安徽省星灵信息科技有限公司 A kind of automatic light concentrator and automatic beam photosystem
CN108275524A (en) * 2018-01-12 2018-07-13 东北大学 A kind of elevator maintenance operation monitoring and guiding device based on the assessment of the first multi-view video series of operations
CN108346144A (en) * 2018-01-30 2018-07-31 哈尔滨工业大学 Bridge Crack based on computer vision monitoring and recognition methods automatically
CN108388641A (en) * 2018-02-27 2018-08-10 广东方纬科技有限公司 A kind of means of transportation based on deep learning ground drawing generating method and system
CN108416283A (en) * 2018-02-28 2018-08-17 华南理工大学 A kind of pavement marking recognition methods based on SSD
CN108416270A (en) * 2018-02-06 2018-08-17 南京信息工程大学 A kind of traffic sign recognition method based on more attribute union features
CN108491889A (en) * 2018-04-02 2018-09-04 深圳市易成自动驾驶技术有限公司 Image, semantic dividing method, device and computer readable storage medium
CN108520286A (en) * 2018-04-24 2018-09-11 青岛科技大学 Infrared dark dim light small target deteection system based on convolutional Neural and candidate region
CN108694829A (en) * 2018-03-27 2018-10-23 西安科技大学 Magnitude of traffic flow identification monitoring network based on unmanned aerial vehicle group mobile platform and method
CN108776834A (en) * 2018-05-07 2018-11-09 上海商汤智能科技有限公司 System enhances learning method and device, electronic equipment, computer storage media
CN108960308A (en) * 2018-06-25 2018-12-07 中国科学院自动化研究所 Traffic sign recognition method, device, car-mounted terminal and vehicle
CN109214441A (en) * 2018-08-23 2019-01-15 桂林电子科技大学 A kind of fine granularity model recognition system and method
CN109492661A (en) * 2018-09-27 2019-03-19 桂林电子科技大学 Traffic sign recognition method based on WWCNN model and TVPDE algorithm
CN109559536A (en) * 2018-12-10 2019-04-02 百度在线网络技术（北京）有限公司 Traffic lights, traffic light recognition method, device, equipment and storage medium
CN109636881A (en) * 2018-12-19 2019-04-16 沈阳天择智能交通工程有限公司 Based on AI identification technology traffic accident situ sketch drafting method
CN109858372A (en) * 2018-12-29 2019-06-07 浙江零跑科技有限公司 A kind of lane class precision automatic Pilot structured data analysis method
CN110163187A (en) * 2019-06-02 2019-08-23 东北石油大学 Remote road traffic sign detection recognition methods based on F-RCNN
CN110188705A (en) * 2019-06-02 2019-08-30 东北石油大学 A kind of remote road traffic sign detection recognition methods suitable for onboard system
CN110312103A (en) * 2019-07-06 2019-10-08 辽宁大学 A kind of high-speed equipment anti-thefting monitoring method for processing video frequency based on cloud computing platform
CN110390344A (en) * 2018-04-19 2019-10-29 华为技术有限公司 Alternative frame update method and device
CN110555877A (en) * 2018-05-31 2019-12-10 杭州海康威视数字技术股份有限公司 Image processing method, device and equipment and readable medium
CN111126571A (en) * 2019-12-25 2020-05-08 福建天晴数码有限公司 R-CNN network optimization method based on DHT network and storage medium
CN111160336A (en) * 2019-12-09 2020-05-15 平安科技（深圳）有限公司 Target detection method, device and computer readable storage medium
CN111222387A (en) * 2018-11-27 2020-06-02 北京嘀嘀无限科技发展有限公司 System and method for object detection
CN111401466A (en) * 2020-03-26 2020-07-10 广州紫为云科技有限公司 Traffic sign detection and identification marking method and device and computer equipment
CN111444976A (en) * 2020-04-02 2020-07-24 Oppo广东移动通信有限公司 Target detection method and device, electronic equipment and readable storage medium
WO2020151153A1 (en) * 2019-01-24 2020-07-30 平安科技（深圳）有限公司 Image processing method and apparatus, and computer device and storage medium
CN111814873A (en) * 2020-07-07 2020-10-23 广州市运通水务有限公司 Method for distinguishing drainage pipeline defect types and automatically identifying defect grades
CN112329497A (en) * 2019-07-18 2021-02-05 杭州海康威视数字技术股份有限公司 Target identification method, device and equipment
CN112784084A (en) * 2019-11-08 2021-05-11 阿里巴巴集团控股有限公司 Image processing method and device and electronic equipment
CN112929665A (en) * 2021-01-28 2021-06-08 北京博雅慧视智能技术研究院有限公司 Target tracking method, device, equipment and medium combining super-resolution and video coding
CN113344880A (en) * 2021-06-09 2021-09-03 浙江国研智能电气有限公司 Fast-RCNN-based low-voltage electrical appliance transfer printing pattern defect detection method
CN114473309A (en) * 2022-03-18 2022-05-13 南京欧睿三维科技有限公司 Welding position identification method for automatic welding system and automatic welding system
CN116901975A (en) * 2023-09-12 2023-10-20 深圳市九洲卓能电气有限公司 Vehicle-mounted AI security monitoring system and method thereof
US11990137B2 (en) 2018-09-13 2024-05-21 Shanghai Cambricon Information Technology Co., Ltd. Image retouching method and terminal device

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20120016461A (en) * 2010-08-16 2012-02-24 주식회사 이미지넥스트 Pavement marking recogniton system and method
CN104766042A (en) * 2014-01-06 2015-07-08 现代摩比斯株式会社 Method and apparatus for and recognizing traffic sign board
CN105930830A (en) * 2016-05-18 2016-09-07 大连理工大学 Road surface traffic sign recognition method based on convolution neural network
CN106295707A (en) * 2016-08-17 2017-01-04 北京小米移动软件有限公司 Image-recognizing method and device
CN106372577A (en) * 2016-08-23 2017-02-01 北京航空航天大学 Deep learning-based traffic sign automatic identifying and marking method
CN106372571A (en) * 2016-08-18 2017-02-01 宁波傲视智绘光电科技有限公司 Road traffic sign detection and identification method
CN106682569A (en) * 2016-09-28 2017-05-17 天津工业大学 Fast traffic signboard recognition method based on convolution neural network

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20120016461A (en) * 2010-08-16 2012-02-24 주식회사 이미지넥스트 Pavement marking recogniton system and method
CN104766042A (en) * 2014-01-06 2015-07-08 现代摩比斯株式会社 Method and apparatus for and recognizing traffic sign board
CN105930830A (en) * 2016-05-18 2016-09-07 大连理工大学 Road surface traffic sign recognition method based on convolution neural network
CN106295707A (en) * 2016-08-17 2017-01-04 北京小米移动软件有限公司 Image-recognizing method and device
CN106372571A (en) * 2016-08-18 2017-02-01 宁波傲视智绘光电科技有限公司 Road traffic sign detection and identification method
CN106372577A (en) * 2016-08-23 2017-02-01 北京航空航天大学 Deep learning-based traffic sign automatic identifying and marking method
CN106682569A (en) * 2016-09-28 2017-05-17 天津工业大学 Fast traffic signboard recognition method based on convolution neural network

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
HOSSEIN P.等: "A new approach in road sign recognition based on fast fractal coding", 《NEURAL COMPUTING AND APPLICATIONS》 *
RONGQIANG Q. 等: "Road Surface Traffic Sign Detection with Hybrid Region Proposal and Fast R-CNN", 《2016 12TH INTERNATIONAL CONFERENCE ON NATURAL COMPUTATION, FUZZY SYSTEMS AND KNOWLEDGE DISCOVERY (ICNC-FSKD)》 *
XUANYUYT: "Fast R-CNN论文详解", 《博客园》 *

Cited By (57)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US12050887B2 (en) 2017-10-30 2024-07-30 Shanghai Cambricon Information Technology Co., Ltd. Information processing method and terminal device
US11922132B2 (en) 2017-10-30 2024-03-05 Shanghai Cambricon Information Technology Co., Ltd. Information processing method and terminal device
US11762631B2 (en) 2017-10-30 2023-09-19 Shanghai Cambricon Information Technology Co., Ltd. Information processing method and terminal device
CN107895191B (en) * 2017-10-30 2022-02-22 上海寒武纪信息科技有限公司 Information processing method and related product
CN107895191A (en) * 2017-10-30 2018-04-10 上海寒武纪信息科技有限公司 A kind of information processing method and Related product
CN107909044A (en) * 2017-11-22 2018-04-13 天津大学 A kind of demographic method of combination convolutional neural networks and trajectory predictions
CN108009581A (en) * 2017-11-30 2018-05-08 中国地质大学（武汉） A kind of method for crack based on CNN, equipment and storage device
CN108095746A (en) * 2017-12-21 2018-06-01 安徽省星灵信息科技有限公司 A kind of automatic light concentrator and automatic beam photosystem
CN108275524A (en) * 2018-01-12 2018-07-13 东北大学 A kind of elevator maintenance operation monitoring and guiding device based on the assessment of the first multi-view video series of operations
CN108346144A (en) * 2018-01-30 2018-07-31 哈尔滨工业大学 Bridge Crack based on computer vision monitoring and recognition methods automatically
CN108346144B (en) * 2018-01-30 2021-03-16 哈尔滨工业大学 Automatic bridge crack monitoring and identifying method based on computer vision
CN108416270A (en) * 2018-02-06 2018-08-17 南京信息工程大学 A kind of traffic sign recognition method based on more attribute union features
CN108416270B (en) * 2018-02-06 2021-07-06 南京信息工程大学 Traffic sign identification method based on multi-attribute combined characteristics
CN108388641A (en) * 2018-02-27 2018-08-10 广东方纬科技有限公司 A kind of means of transportation based on deep learning ground drawing generating method and system
CN108388641B (en) * 2018-02-27 2022-02-01 广东方纬科技有限公司 Traffic facility map generation method and system based on deep learning
CN108416283A (en) * 2018-02-28 2018-08-17 华南理工大学 A kind of pavement marking recognition methods based on SSD
CN108694829B (en) * 2018-03-27 2021-08-13 西安科技大学 Traffic flow identification monitoring network and method based on unmanned aerial vehicle group mobile platform
CN108694829A (en) * 2018-03-27 2018-10-23 西安科技大学 Magnitude of traffic flow identification monitoring network based on unmanned aerial vehicle group mobile platform and method
CN108491889A (en) * 2018-04-02 2018-09-04 深圳市易成自动驾驶技术有限公司 Image, semantic dividing method, device and computer readable storage medium
CN110390344A (en) * 2018-04-19 2019-10-29 华为技术有限公司 Alternative frame update method and device
CN108520286A (en) * 2018-04-24 2018-09-11 青岛科技大学 Infrared dark dim light small target deteection system based on convolutional Neural and candidate region
CN108776834A (en) * 2018-05-07 2018-11-09 上海商汤智能科技有限公司 System enhances learning method and device, electronic equipment, computer storage media
US11669711B2 (en) 2018-05-07 2023-06-06 Shanghai Sensetime Intelligent Technology Co., Ltd System reinforcement learning method and apparatus, and computer storage medium
CN110555877A (en) * 2018-05-31 2019-12-10 杭州海康威视数字技术股份有限公司 Image processing method, device and equipment and readable medium
CN108960308A (en) * 2018-06-25 2018-12-07 中国科学院自动化研究所 Traffic sign recognition method, device, car-mounted terminal and vehicle
CN109214441A (en) * 2018-08-23 2019-01-15 桂林电子科技大学 A kind of fine granularity model recognition system and method
US12057109B2 (en) 2018-09-13 2024-08-06 Shanghai Cambricon Information Technology Co., Ltd. Information processing method and terminal device
US11990137B2 (en) 2018-09-13 2024-05-21 Shanghai Cambricon Information Technology Co., Ltd. Image retouching method and terminal device
US11996105B2 (en) 2018-09-13 2024-05-28 Shanghai Cambricon Information Technology Co., Ltd. Information processing method and terminal device
US12057110B2 (en) 2018-09-13 2024-08-06 Shanghai Cambricon Information Technology Co., Ltd. Voice recognition based on neural networks
US12094456B2 (en) 2018-09-13 2024-09-17 Shanghai Cambricon Information Technology Co., Ltd. Information processing method and system
CN109492661B (en) * 2018-09-27 2021-08-13 桂林电子科技大学 Traffic sign identification method based on WWCNN model and TVPDE algorithm
CN109492661A (en) * 2018-09-27 2019-03-19 桂林电子科技大学 Traffic sign recognition method based on WWCNN model and TVPDE algorithm
CN111222387A (en) * 2018-11-27 2020-06-02 北京嘀嘀无限科技发展有限公司 System and method for object detection
CN111222387B (en) * 2018-11-27 2023-03-03 北京嘀嘀无限科技发展有限公司 System and method for object detection
WO2020107510A1 (en) * 2018-11-27 2020-06-04 Beijing Didi Infinity Technology And Development Co., Ltd. Ai systems and methods for objection detection
CN109559536A (en) * 2018-12-10 2019-04-02 百度在线网络技术（北京）有限公司 Traffic lights, traffic light recognition method, device, equipment and storage medium
CN109636881A (en) * 2018-12-19 2019-04-16 沈阳天择智能交通工程有限公司 Based on AI identification technology traffic accident situ sketch drafting method
CN109858372A (en) * 2018-12-29 2019-06-07 浙江零跑科技有限公司 A kind of lane class precision automatic Pilot structured data analysis method
WO2020151153A1 (en) * 2019-01-24 2020-07-30 平安科技（深圳）有限公司 Image processing method and apparatus, and computer device and storage medium
CN110188705B (en) * 2019-06-02 2022-05-06 东北石油大学 Remote traffic sign detection and identification method suitable for vehicle-mounted system
CN110163187A (en) * 2019-06-02 2019-08-23 东北石油大学 Remote road traffic sign detection recognition methods based on F-RCNN
CN110188705A (en) * 2019-06-02 2019-08-30 东北石油大学 A kind of remote road traffic sign detection recognition methods suitable for onboard system
CN110312103A (en) * 2019-07-06 2019-10-08 辽宁大学 A kind of high-speed equipment anti-thefting monitoring method for processing video frequency based on cloud computing platform
CN112329497A (en) * 2019-07-18 2021-02-05 杭州海康威视数字技术股份有限公司 Target identification method, device and equipment
CN112784084B (en) * 2019-11-08 2024-01-26 阿里巴巴集团控股有限公司 Image processing method and device and electronic equipment
CN112784084A (en) * 2019-11-08 2021-05-11 阿里巴巴集团控股有限公司 Image processing method and device and electronic equipment
CN111160336A (en) * 2019-12-09 2020-05-15 平安科技（深圳）有限公司 Target detection method, device and computer readable storage medium
CN111126571A (en) * 2019-12-25 2020-05-08 福建天晴数码有限公司 R-CNN network optimization method based on DHT network and storage medium
CN111401466A (en) * 2020-03-26 2020-07-10 广州紫为云科技有限公司 Traffic sign detection and identification marking method and device and computer equipment
CN111444976A (en) * 2020-04-02 2020-07-24 Oppo广东移动通信有限公司 Target detection method and device, electronic equipment and readable storage medium
CN111814873A (en) * 2020-07-07 2020-10-23 广州市运通水务有限公司 Method for distinguishing drainage pipeline defect types and automatically identifying defect grades
CN112929665A (en) * 2021-01-28 2021-06-08 北京博雅慧视智能技术研究院有限公司 Target tracking method, device, equipment and medium combining super-resolution and video coding
CN113344880A (en) * 2021-06-09 2021-09-03 浙江国研智能电气有限公司 Fast-RCNN-based low-voltage electrical appliance transfer printing pattern defect detection method
CN114473309A (en) * 2022-03-18 2022-05-13 南京欧睿三维科技有限公司 Welding position identification method for automatic welding system and automatic welding system
CN116901975A (en) * 2023-09-12 2023-10-20 深圳市九洲卓能电气有限公司 Vehicle-mounted AI security monitoring system and method thereof
CN116901975B (en) * 2023-09-12 2023-11-21 深圳市九洲卓能电气有限公司 Vehicle-mounted AI security monitoring system and method thereof

Also Published As

Publication number Publication date
CN107301383B (en) 2020-11-24

Similar Documents

Publication Publication Date Title
CN107301383A (en) A kind of pavement marking recognition methods based on Fast R CNN
CN110942000B (en) Unmanned vehicle target detection method based on deep learning
CN108416283A (en) A kind of pavement marking recognition methods based on SSD
Bi et al. Improved VGG model-based efficient traffic sign recognition for safe driving in 5G scenarios
CN104850845B (en) A kind of traffic sign recognition method based on asymmetric convolutional neural networks
CN113486726A (en) Rail transit obstacle detection method based on improved convolutional neural network
CN111612807A (en) Small target image segmentation method based on scale and edge information
CN106709568A (en) RGB-D image object detection and semantic segmentation method based on deep convolution network
CN105260712A (en) Method and system for detecting pedestrian in front of vehicle
Guindel et al. Fast joint object detection and viewpoint estimation for traffic scene understanding
Kanaeva et al. Road pavement crack detection using deep learning with synthetic data
CN113159215A (en) Small target detection and identification method based on fast Rcnn
CN114973199A (en) Rail transit train obstacle detection method based on convolutional neural network
Lian et al. Towards unified on-road object detection and depth estimation from a single image
Guan et al. Multi-scale object detection with feature fusion and region objectness network
CN117975167B (en) Weak ore spot ore sorting method, device, readable storage medium and equipment
CN116740516A (en) Target detection method and system based on multi-scale fusion feature extraction
Das et al. Deep multi-task networks for occluded pedestrian pose estimation
Zhang et al. Bus passenger flow statistics algorithm based on deep learning
Zhang et al. Pre-locate net for object detection in high-resolution images
Wu et al. Compositional Oil Spill Detection Based on Object Detector and Adapted Segment Anything Model from SAR Images
Kim et al. Domain-specific data augmentation for on-road object detection based on a deep neural network
Huo et al. Traffic sign recognition based on resnet-20 and deep mutual learning
Nguyen et al. Robust vehicle detection under adverse weather conditions using auto-encoder feature
Liu et al. Research on pedestrian detection algorithm in driverless urban traffic environment

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant