CN106446864A - Method for detecting feasible road - Google Patents
Method for detecting feasible road Download PDFInfo
- Publication number
- CN106446864A CN106446864A CN201610888826.5A CN201610888826A CN106446864A CN 106446864 A CN106446864 A CN 106446864A CN 201610888826 A CN201610888826 A CN 201610888826A CN 106446864 A CN106446864 A CN 106446864A
- Authority
- CN
- China
- Prior art keywords
- feature
- road
- connecting way
- training
- pixel
- 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
- G06F18/21—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
- 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
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a method for detecting a feasible road. The method comprises the following steps of a training phase and a testing phase; in the testing phase, extracting feature information of a whole image, such as color, gradient, direction and boundary; then, primarily detecting the feasible road by a classifier; utilizing the primary road detection result to establish two-dimensional color histograms of the road area and the non-road area, calculating the road probability of each pixel point, and utilizing the road probability to update the detection result of the feasible road. The method has the advantages that any preset assumption is not needed, the suitability is strong, and the method is suitable for various scenes; by not utilizing the image division method, the calculation amount is greatly reduced, the real-time processing requirement is met, and the feasible road area is accurately detected.
Description
Technical field
The present invention relates to the detection method of a kind of connecting way, particularly relate to one be applicable to road traffic detection,
The detection method of connecting way.
Background technology
In recent years, day by day complicated with the increase of Vehicle's quantity and road traffic condition, traffic safety problem is got over
More paid attention to the people by country, the field such as the Intelligent unattended driving that thus excites, unmanned plane, intelligent mobile robot
Technical research becomes more and more hotter, and lane detection technology is one of key point in these technical research.Lane detection technology is to respectively
Plant road scene and carry out perception and understanding, testing result is exported to navigation system or path planning system, in order to instruct mobile
Equipment next step should be taked which kind of action.Therefore, the quality of Road Detection result will directly affect mobile device independent navigation
Accuracy.
Mainly there are road detection and road area detection, road detection one for the research of lane detection technology at present
As be applied in the DAS (Driver Assistant System) under highway scene, the mainly road marking line of white or yellow on detection highway, pin
Stronger to property, it is impossible to extensively to apply；Road area detects applied range, and its technology is broadly divided into two big classes：Based on road
The matching method of model and the characteristic method based on roadway characteristic.Matching method sets up road model according to the priori of road, utilizes
Model parameter carries out path adaptation, and the method can effectively overcome road surface to pollute, shade and the environmental factor such as uneven illumination is even, but
When road does not meets and pre-supposes that, model will lose efficacy.Characteristic method is special at image with non-rice habitats region mainly by road
The difference levied carries out Road Detection, and these features include shape, gray scale, texture and contrast etc., and reasonable method has base
In the Road Detection of super-pixel segmentation, but the method amount of calculation is bigger, is extremely difficult to examine in real time on some embedded platforms
The requirement surveyed.
Content of the invention
The technical problem to be solved in the present invention is to provide a kind of based on characteristics of image, the detection method of connecting way, should
Method is not based on any pre-supposing that, applicability is strong, and operand is low, disclosure satisfy that the requirement of process in real time, simultaneously can also be accurate
Detect connecting way region.
The technical solution used in the present invention is as follows：The detection method of a kind of connecting way, concrete grammar is：Including training rank
Section and test phase；Entire image is first extracted the characteristic informations such as color, gradient, direction and edge by test phase, then passes through
Grader tentatively carries out connecting way detection；Utilize preliminary Road Detection result, set up road area and non-rice habitats region
Two-dimensional color histogram, calculates the road probability of each pixel, then utilizes the testing result of road probability updating connecting way.
The method is not based on any pre-supposing that, it is possible to adapt to various scene.In addition the method does not utilize image partition method,
Greatly reduce amount of calculation, meet the requirement of process in real time, connecting way region can also be accurately detected out simultaneously.
Described method also includes：At test phase, finally by detection barrier, again to road area and non-rice habitats district
Territory is updated, and generates accurate connecting way testing result.
Concrete grammar step is：
Training stage：
S11, set up training sample database：Collect the picture comprising road in a large number, be divided into training pictures and test pictures,
And the road in each pictures is labeled；
S12, foundation training grader：A given picture comprising road, utilizes markup information, from given figure
Intercept multiple picture comprising road in Xiang and put into positive Sample Storehouse, and intercept and multiple comprise off-highroad picture and put into negative sample
Storehouse, thus set up and complete positive negative example base；
S13, the every piece image reading in positive negative example base, go out YUV channel information by RGB color path computation, so
Afterwards to Y path computation gradient magnitude GM, gradient direction GO and edge E；Y feature is generated for Y passage, then for U passage, V
Passage, gradient magnitude GM and edge E, carry out the operation of Y passage equally, generate the U feature of dimension as Y feature, V feature,
GM feature and E feature；For gradient direction GO, generate the GO feature of 6 dimensions as Y feature；Utilize the same of above generation
The Y feature of sample dimension, U feature, V feature, GM feature, E feature and GO features training can detect the connecting way district in picture
The Adaboost in territory or SVM classifier；
Test phase：
The grader Preliminary detection connecting way that S21, utilization train, concrete grammar is：Input picture is divided into one
Fixation element fritter, does not has overlapping region between each fritter；Each fritter is extracted the same dimension described in S13 Y feature,
U feature, V feature, GM feature, E feature and GO feature, and input the feature in the Adaboost training or SVM classifier,
Judge whether this fritter belongs to road area；
S22, utilize two-dimensional color histogram update testing result, concrete grammar is：To the road area detecting in S21
Calculate the two-dimensional histogram of Y and U with non-rice habitats area distribution, then calculate the probability that every a pair Y and U belongs to road, utilize this
Whether each pixel that probability rejudges image belongs to road area.
Described method step also includes：Again updating testing result while S22, detection barrier, concrete grammar is：?
On the basis of two-dimensional color histogram updates testing result, to each non-rice habitats area pixel point, calculate it up and down
Distance with road area pixel on four direction, if four distances exist and both less than set threshold value T, then it is assumed that this is non-
Road area pixel is road pixel point, otherwise it is assumed that this non-rice habitats area pixel point is barrier pixel.By this kind
Mode, further have updated the testing result in connecting way region, makes result more accurate while detection barrier.
Described S22 also includes：On the basis of two-dimensional color histogram updates testing result, to each road area picture
Vegetarian refreshments, calculates its distance with non-rice habitats area pixel point on four direction up and down, if four distance exist and all
Less than setting threshold value T, then it is assumed that this road area pixel is non-rice habitats pixel, otherwise it is assumed that this road area pixel is
Barrier pixel.By this kind of mode, while detection barrier, further have updated the detection knot in connecting way region
Really, make result more accurate.
In described S11, when being labeled the road in each pictures, enter rower along the border in connecting way region
Note.
In described S11, when classifying training pictures and test pictures, randomly select the sample of half as instruction
Practice pictures, remaining as test pictures.
In described S13, Y feature, U feature, V feature, GM feature, E feature and GO feature are the feature of 64 dimensions.
In the training stage, also include S14, use test data to test grader, by misclassification in test sample
Sample is elected, and joins in test sample, and re-training grader improves the degree of accuracy with this.
Compared with prior art, the invention has the beneficial effects as follows：Being not based on any pre-supposing that, applicability is strong, Neng Goushi
Close various scene；Do not utilize image partition method, greatly reduce amount of calculation, disclosure satisfy that the requirement of process in real time, simultaneously
Connecting way region can also be accurately detected out.
Brief description
Fig. 1 is the principle schematic of a present invention wherein embodiment.
Detailed description of the invention
In order to make the purpose of the present invention, technical scheme and advantage clearer, below in conjunction with drawings and Examples, right
The present invention is further elaborated.It should be appreciated that specific embodiment described herein only in order to explain the present invention, not
For limiting the present invention.
Any feature disclosed in this specification (including summary and accompanying drawing), unless specifically stated otherwise, all can be by other equivalences
Or the alternative features with similar purpose is replaced.I.e., unless specifically stated otherwise, each feature is a series of equivalence or class
Like one of feature example.
Specific embodiment 1
The detection method of a kind of connecting way, concrete grammar is：Including training stage and test phase；Test phase is first right
Entire image extracts the characteristic informations such as color, gradient, direction and edge, is then tentatively carried out connecting way inspection by grader
Survey；Utilize preliminary Road Detection result, set up the two-dimensional color histogram in road area and non-rice habitats region, calculate each pixel
The road probability of point, then utilizes the testing result of road probability updating connecting way.The method is not based on any pre-supposing that,
So being adapted to various scene.In addition the method does not utilize image partition method, greatly reduces amount of calculation, meets reality
When the requirement that processes, simultaneously can also accurately detect out connecting way region.
Specific embodiment 2
On the basis of being embodied as 1, described method also includes：At test phase, finally by detection barrier, again
Road area and non-rice habitats region are updated, generate accurate connecting way testing result.
Specific embodiment 3
As it is shown in figure 1, on the basis of specific embodiment 1 or 2, concrete grammar step is：
Training stage：
S11, set up training sample database：Collect the picture comprising road in a large number, be divided into training pictures and test pictures,
And the road in each pictures is labeled；In this specific embodiment, by the road mark in picture
GroundTruth；
S12, foundation training grader：A given picture comprising road, utilizes markup information, from given figure
Intercept multiple picture comprising road in Xiang and put into positive Sample Storehouse, and intercept and multiple comprise off-highroad picture and put into negative sample
Storehouse, thus set up and complete positive negative example base；In this specific embodiment, the picture intercepting from given image is 128 × 128
Square picture, and carry out even partition；
S13, the every piece image reading in positive negative example base, go out YUV channel information by RGB color path computation, so
Afterwards to Y path computation gradient magnitude GM, gradient direction GO and edge E；Y feature is generated for Y passage, then for U passage, V
Passage, gradient magnitude GM and edge E, carry out the operation of Y passage equally, generate the U feature of dimension as Y feature, V feature,
GM feature and E feature；For gradient direction GO, generate the GO feature of 6 dimensions as Y feature；Utilize the same of above generation
The Y feature of sample dimension, U feature, V feature, GM feature, E feature and GO features training can detect the connecting way district in picture
The Adaboost in territory or SVM classifier；In this specific embodiment, each width picture reading is downsampled to 32 × 32；Right
In Y channel image, ask for all elements inside 4x4 block and as one-dimensional characteristic：
Wherein YijFor j-th pixel of i-th 4 × 4 pieces of the insides in Y channel image, fYiFor i-th dimension feature, wherein i
=1,2 ..., 64；J=1,2 ..., 16；So Y channel image can generate 64 dimension Y features：
FY=[fY1,fY2,…,fY64,] (2)
For U passage, V passage, gradient magnitude GM and edge E, carry out the operation of Y passage equally, similarly generate feature dimensions
Degree is U feature, V feature, GM feature and the E feature of 64 dimensions；For gradient direction GO, in 4 × 4 pieces of the insides, 6 directions of statistics
Amplitude accumulation information as the one-dimensional characteristic in each direction, such as calculate it at θkOne-dimensional characteristic on direction：
Wherein θkFor quantify gradient direction, k=1,2 ..., 6.GMijAnd GOijRepresent inside GM and GO image i-th respectively
The corresponding value of j-th pixel of individual 4 × 4 pieces of the insides, i=1,2 ..., 64；J=1,2 ..., 16.
So at θkThe feature of 64 dimensions just can be generated on direction：
So gradient direction just can generate the GO feature of 6 64 dimensions：
FGO=[Fθ1,Fθ2,Fθ3,Fθ4,Fθ5,Fθ6] (5)
Stringing together features above, final piece image can be taken off 704 dimensional features：
F=[Fy,FU,FV,FGM,FE,FGO] (6)
Utilizing features training Adaboost extracted from positive negative example base or SVM classifier, this grader can be rough
Connecting way region in detection picture；
Test phase：
The grader Preliminary detection connecting way that S21, utilization train, concrete grammar is：Input picture is divided into one
Fixation element fritter, does not has overlapping region between each fritter；Each fritter is extracted the same dimension described in S13 Y feature,
U feature, V feature, GM feature, E feature and GO feature, and input the feature in the Adaboost training or SVM classifier,
Judge whether this fritter belongs to road area；In this specific embodiment, input picture is evenly divided into the fritter of 32 × 32,
Not having overlapping region between each fritter, unnecessary pixel does not processes.Training stage institute is extracted to each fritter of 32 × 32
704 dimensional features stated, and input the feature in the Adaboost training or SVM classifier, it is judged that whether this fritter belongs to
Road area, so can detect the connecting way region in image roughly；
S22, utilize two-dimensional color histogram update testing result, concrete grammar is：To the road area detecting in S21
Calculate the two-dimensional histogram of Y and U with non-rice habitats area distribution
P (Y, U)=nY,U/N (7)
Wherein nY, URepresenting the number that in road area or non-rice habitats region, pixel point value is (Y, U), N represents road area
Or the sum of pixel in non-rice habitats region.Wherein Y=0,1 ..., 255；U=0,1 ..., 255.Calculate every a pair Y and U again
Belong to the probability of road
Wherein pr(Y, U) represents the probability of a pair Y and U appearance in road area, pu(Y, U) represents in non-rice habitats region one
The probability that Y and U is occurred.
Utilize this probability (Proad(Y, U)) whether each pixel of rejudging image belong to road area, when this
The corresponding probability of pixel is more than 0.5, then it is assumed that be road pixel point, is otherwise non-rice habitats pixel.This makes it possible to obtain
Connecting way region more accurately.
In this specific embodiment, GroundTruth markup information according to picture, comprise road in calculating each square
The percentage in region.If this ratio is higher than 80%, thinking that this square belongs to road block, positive Sample Storehouse is out put in cutting；If should
Less than 10%, ratio thinks that this square belongs to non-rice habitats block, negative example base is out put in cutting.
Specific embodiment 4
As it is shown in figure 1, on the basis of specific embodiment 3, described method step also includes：S23, detection barrier same
Shi Zaici updates testing result, and concrete grammar is：On the basis of two-dimensional color histogram updates testing result, non-to each
Road area pixel, calculates its distance with road area pixel on four direction up and down, if four distances
Exist and both less than set threshold value T, then it is assumed that this non-rice habitats area pixel point is road pixel point, otherwise it is assumed that this non-rice habitats district
Territory pixel is barrier pixel.By this kind of mode, while detection barrier, further have updated connecting way district
The testing result in territory, makes result more accurate.
Specific embodiment 5
On the basis of specific embodiment 4, described S23 also includes：Update the base of testing result at two-dimensional color histogram
On plinth, to each road area pixel, calculate its on four direction up and down with non-rice habitats area pixel point away from
From if four distances exist and both less than set threshold value T, then it is assumed that this road area pixel is non-rice habitats pixel, no
Then think that this road area pixel is barrier pixel.By this kind of mode, while detection barrier further more
The new testing result in connecting way region, makes result more accurate.
Specific embodiment 6
On the basis of one of specific embodiment 3 to 5, in described S11, the road in each pictures is labeled
When, it is labeled along the border in connecting way region.
Specific embodiment 7
On the basis of one of specific embodiment 3 to 6, in described S11, training pictures and test pictures are carried out point
During class, randomly select the sample of half as training pictures, remaining conduct test pictures.
Specific embodiment 8
On the basis of one of specific embodiment 1 to 7, in the training stage, also include S14, use test data to classification
Device is tested, and elects the sample of misclassification in test sample, joins in test sample, and re-training grader, with this
Improve the degree of accuracy.
Claims (9)
1. a detection method for connecting way, concrete grammar is：Including training stage and test phase；Test phase is first to whole
The characteristic informations such as width image zooming-out color, gradient, direction and edge, are then tentatively carried out connecting way detection by grader；
Utilize preliminary Road Detection result, set up the two-dimensional color histogram in road area and non-rice habitats region, calculate each pixel
Road probability, then utilize the testing result of road probability updating connecting way.
2. the detection method of connecting way according to claim 1, described method also includes：At test phase, finally lead to
Cross detection barrier, again road area and non-rice habitats region are updated, generate accurate connecting way testing result.
3. the detection method of connecting way according to claim 1 and 2, concrete grammar step is：
Training stage：
S11, set up training sample database：Collect the picture comprising road in a large number, be divided into training pictures and test pictures, and will
Road in each pictures is labeled；
S12, foundation training grader：A given picture comprising road, utilizes markup information, from given image
Intercept multiple picture comprising road and put into positive Sample Storehouse, and intercept and multiple comprise off-highroad picture and put into negative example base, from
And set up and complete positive negative example base；
S13, the every piece image reading in positive negative example base, go out YUV channel information by RGB color path computation, then to Y
Path computation gradient magnitude GM, gradient direction GO and edge E；For Y passage generate Y feature, then for U passage, V passage,
Gradient magnitude GM and edge E, carries out the operation of Y passage, the U feature of generation dimension as Y feature, V feature, GM feature equally
With E feature；For gradient direction GO, generate the GO feature of 6 dimensions as Y feature；Utilize the above same dimension generating
Y feature, U feature, V feature, GM feature, E feature and GO features training can detect connecting way region in picture
Adaboost or SVM classifier；
Test phase：
The grader Preliminary detection connecting way that S21, utilization train, concrete grammar is：Input picture is divided into a fixation
Element fritter, does not has overlapping region between each fritter；Extract the Y feature of the same dimension described in S13, U spy to each fritter
Levy, V feature, GM feature, E feature and GO feature, and input the feature in the Adaboost training or SVM classifier, sentence
Whether this fritter disconnected belongs to road area；
S22, utilize two-dimensional color histogram update testing result, concrete grammar is：To the road area detecting in S21 and non-
Road area distribution calculates the two-dimensional histogram of Y and U, then calculates the probability that every a pair Y and U belongs to road, utilizes this probability
Whether each pixel rejudging image belongs to road area.
4. the detection method of connecting way according to claim 3, described method step also includes：S22, detection barrier
While again update testing result, concrete grammar is：On the basis of two-dimensional color histogram updates testing result, to each
Individual non-rice habitats area pixel point, calculates its distance with road area pixel on four direction up and down, if four
Distance exists and both less than sets threshold value T, then it is assumed that this non-rice habitats area pixel point is road pixel point, otherwise it is assumed that this non-road
Road area pixel point is barrier pixel.
5. the detection method of connecting way according to claim 4, described S22 also includes：At two-dimensional color histogram more
On the basis of new testing result, to each road area pixel, calculate its on four direction up and down with non-rice habitats
The distance of area pixel point, if four distances exist and both less than set threshold value T, then it is assumed that this road area pixel is non-
Road pixel point, otherwise it is assumed that this road area pixel is barrier pixel.
6. the detection method of connecting way according to claim 3, in described S11, enters to the road in each pictures
It during rower note, is labeled along the border in connecting way region.
7. the detection method of connecting way according to claim 3, in described S11, to training pictures and test picture
When collection is classified, randomly select the sample of half as training pictures, remaining conduct test pictures.
8. the detection method of connecting way according to claim 3, in described S13, Y feature, U feature, V feature, GM are special
Levy, E feature and GO feature are the feature of 64 dimensions.
9. the detection method of connecting way according to claim 3, in the training stage, also includes S14, uses test data
Testing grader, electing the sample of misclassification in test sample, join in test sample, re-training is classified
Device.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610888826.5A CN106446864A (en) | 2016-10-12 | 2016-10-12 | Method for detecting feasible road |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610888826.5A CN106446864A (en) | 2016-10-12 | 2016-10-12 | Method for detecting feasible road |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106446864A true CN106446864A (en) | 2017-02-22 |
Family
ID=58174066
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610888826.5A Pending CN106446864A (en) | 2016-10-12 | 2016-10-12 | Method for detecting feasible road |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106446864A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107901907A (en) * | 2017-09-30 | 2018-04-13 | 惠州市德赛西威汽车电子股份有限公司 | A kind of method for detecting lane lines based on color lump statistics |
| CN108052904A (en) * | 2017-12-13 | 2018-05-18 | 辽宁工业大学 | The acquisition methods and device of lane line |
| CN108197569A (en) * | 2017-12-29 | 2018-06-22 | 驭势科技（北京）有限公司 | Obstacle recognition method, device, computer storage media and electronic equipment |
| CN113139975A (en) * | 2021-04-19 | 2021-07-20 | 国交空间信息技术（北京）有限公司 | Road characteristic-based road surface segmentation method and device |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100017060A1 (en) * | 2008-07-18 | 2010-01-21 | Gm Global Technology Operations, Inc. | Road-edge detection |
| CN105095869A (en) * | 2015-07-24 | 2015-11-25 | 深圳市佳信捷技术股份有限公司 | Pedestrian detection method and apparatus |
| CN105335716A (en) * | 2015-10-29 | 2016-02-17 | 北京工业大学 | Improved UDN joint-feature extraction-based pedestrian detection method |
| CN105989334A (en) * | 2015-02-12 | 2016-10-05 | 中国科学院西安光学精密机械研究所 | Monocular vision-based road detection method |
-
2016
- 2016-10-12 CN CN201610888826.5A patent/CN106446864A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100017060A1 (en) * | 2008-07-18 | 2010-01-21 | Gm Global Technology Operations, Inc. | Road-edge detection |
| CN105989334A (en) * | 2015-02-12 | 2016-10-05 | 中国科学院西安光学精密机械研究所 | Monocular vision-based road detection method |
| CN105095869A (en) * | 2015-07-24 | 2015-11-25 | 深圳市佳信捷技术股份有限公司 | Pedestrian detection method and apparatus |
| CN105335716A (en) * | 2015-10-29 | 2016-02-17 | 北京工业大学 | Improved UDN joint-feature extraction-based pedestrian detection method |
Non-Patent Citations (2)
| Title |
|---|
| CERYAN TAN，ET AL.: "Color Model-Based Real-Time Learning for Road Following", 《2006 IEEE INTELLIGENT TRANSPORTATION SYSTEMS CONFERENCE》 * |
| 刘昶: "基于YUV色彩空间的车辆视频检测算法及实现", 《中国优秀博硕士学位论文全文数据库 (硕士) 信息科技辑》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107901907A (en) * | 2017-09-30 | 2018-04-13 | 惠州市德赛西威汽车电子股份有限公司 | A kind of method for detecting lane lines based on color lump statistics |
| CN107901907B (en) * | 2017-09-30 | 2019-12-20 | 惠州市德赛西威汽车电子股份有限公司 | Lane line detection method based on color block statistics |
| CN108052904A (en) * | 2017-12-13 | 2018-05-18 | 辽宁工业大学 | The acquisition methods and device of lane line |
| CN108052904B (en) * | 2017-12-13 | 2021-11-30 | 辽宁工业大学 | Method and device for acquiring lane line |
| CN108197569A (en) * | 2017-12-29 | 2018-06-22 | 驭势科技（北京）有限公司 | Obstacle recognition method, device, computer storage media and electronic equipment |
| CN113139975A (en) * | 2021-04-19 | 2021-07-20 | 国交空间信息技术（北京）有限公司 | Road characteristic-based road surface segmentation method and device |
| CN113139975B (en) * | 2021-04-19 | 2023-11-17 | 国交空间信息技术（北京）有限公司 | Road feature-based pavement segmentation method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110175576B (en) | Driving vehicle visual detection method combining laser point cloud data | |
| CN105260699B (en) | A kind of processing method and processing device of lane line data | |
| Kong et al. | General road detection from a single image | |
| Danescu et al. | Detection and classification of painted road objects for intersection assistance applications | |
| CN106682586A (en) | Method for real-time lane line detection based on vision under complex lighting conditions | |
| CN105488454A (en) | Monocular vision based front vehicle detection and ranging method | |
| CN105160309A (en) | Three-lane detection method based on image morphological segmentation and region growing | |
| CN103871077B (en) | A kind of extraction method of key frame in road vehicles monitoring video | |
| Huang et al. | Spatial-temproal based lane detection using deep learning | |
| CN103679168A (en) | Detection method and detection device for character region | |
| CN114998852A (en) | Intelligent detection method for road pavement diseases based on deep learning | |
| CN103605953A (en) | Vehicle interest target detection method based on sliding window search | |
| CN102867183B (en) | Method and device for detecting littered objects of vehicle and intelligent traffic monitoring system | |
| CN106446864A (en) | Method for detecting feasible road | |
| CN104050684B (en) | A kind of video frequency motion target sorting technique based on on-line training and system | |
| CN107767400A (en) | Remote sensing images sequence moving target detection method based on stratification significance analysis | |
| CN106446785A (en) | Passable road detection method based on binocular vision | |
| CN110096981A (en) | A kind of video big data traffic scene analysis method based on deep learning | |
| CN102915433A (en) | Character combination-based license plate positioning and identifying method | |
| CN103679205A (en) | Preceding car detection method based on shadow hypothesis and layered HOG (histogram of oriented gradient) symmetric characteristic verification | |
| CN105005989A (en) | Vehicle target segmentation method under weak contrast | |
| CN105069774A (en) | Object segmentation method based on multiple-instance learning and graph cuts optimization | |
| CN106980855A (en) | Traffic sign quickly recognizes alignment system and method | |
| CN103021183A (en) | Method for detecting regulation-violating motor vehicles in monitoring scene | |
| CN108629297A (en) | A kind of remote sensing images cloud detection method of optic based on spatial domain natural scene statistics |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into 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: 20170222 |