CN110458023A - A kind of training method of road detection model, road detection method and device - Google Patents
A kind of training method of road detection model, road detection method and device Download PDFInfo
- Publication number
- CN110458023A CN110458023A CN201910624027.0A CN201910624027A CN110458023A CN 110458023 A CN110458023 A CN 110458023A CN 201910624027 A CN201910624027 A CN 201910624027A CN 110458023 A CN110458023 A CN 110458023A
- Authority
- CN
- China
- Prior art keywords
- road
- detection model
- characteristic pattern
- training
- prediction
- 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
Links
Classifications
-
- 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
-
- 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
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 discloses a kind of training methods of road detection model, road detection method and device, are related to field of image processing, further relate to intelligent driving field.Using the training method of road detection model disclosed by the invention, using the continuous spatial position continuity of linear operator extraction characteristic pattern and the linear shape of Road, supervised training obtains depth Road detection model, multiframe integration technology generates prediction road information according to vehicle movement track and history road information, according to the part being blocked in prediction Road correction present road line, the robustness of detection is strengthened, so that road detection model is more suitable for applying in actual condition.
Description
Technical field
The present invention relates to field of image processings, further relate to intelligent driving field more particularly to a kind of road detection model
Training method, road detection method and device.
Background technique
In recent years, intellectualizing system is widely used in driving field, it is therefore intended that realizes Function for Automatic Pilot or auxiliary
Drive function.In the visual perception system of intelligent driving, lane detection is an important functional module, it usually needs from vehicle
Lane line is detected in road image around, and then instructs to drive vehicle.
Currently, academia and industry develop many deep learning methods for detection lane line and lane boundary line,
Wherein, the representative are LaneNet and SCNN.LaneNet is quoted from disclosed in IEEE IV conference in 2018
Paper " Towards End-to-End Lane Detection:an Instance Segmentation Approach ".It should
The flow chart of method as shown in Figure 1, LaneNet has trained Liang Ge branch, by the pixel to image carry out lane line point and
It after the semantic segmentation of non-lane line point, then birdss of the same feather flock together to segmentation result, obtains the pixel set for belonging to several lane lines.
SCNN is quoted from the paper " Spatial disclosed in AAAI Conference on Artificial Intelligence in 2018
As Deep:Spatial CNN for Traffic Scene Understanding".The flow chart of this method as shown in Fig. 2,
SCNN passes through the feature that convolutional Neural net extracts image first, then using Spatial CNN to the horizontal and vertical continuous of image
Property is modeled.
Above two method achieves preferable detection effect, but there is also respective defects in actual condition.For example,
When using LaneNet method, it is easy to get desultory testing result, when using SCNN, is easy the continuous spot on road surface
Equal erroneous detections are lane line or lane boundary line.And it is not suitable for lane line or the case where lane boundary line is blocked, and in reality
In the application scenarios of border, the case where lane line and lane boundary line are blocked, is very common, using any one in the above method
Exist to testing result and interferes.
Summary of the invention
Technical problem to be solved by the present invention lies in provide training method, the Road of a kind of road detection model
Detection method and device solve the problems, such as that the prior art can not accurately detect lane line and highway boundary line and detect effect
The bad problem of fruit.
In view of this, the present invention provides a kind of training method of road detection model, the road detection model
Training method includes:
Road sample image is obtained,
Specifically, the road sample image includes at least the Road data of mark.
The feature of the road sample image is extracted by convolutional network, obtains characteristic pattern.
Using the further feature of characteristic pattern described in continuous linear operator extraction,
Specifically, the further feature includes the spatial position continuity of Road and the linear form of Road.
According to the further feature, is exercised supervision training with the Road data of mark, optimize deep learning network losses
Function adjusts network parameter, obtains road detection model.
Further, include: using the step of further feature of characteristic pattern described in continuous linear operator extraction
Convolution is iterated to the horizontal and vertical of the characteristic pattern using convolution kernel,
Specifically, described horizontal and vertical including from right to left, from left to right, from top to bottom and from top to bottom.
Obtain the point probability value of the characteristic pattern in one direction.
According to described probability value, judges the characteristic pattern in this direction and whether there is linear form,
If it exists, then the further feature of the characteristic pattern is extracted；
If it does not exist, then inhibit the direction that linear form is not present.
Further, the step of judgement characteristic pattern whether there is linear form in this direction are as follows:
Whether the sum of the point probability value of judgement in this direction is greater than first threshold；
If more than then judging whether the difference of point probability value in this direction is less than second threshold；
If being less than, determine that the characteristic pattern has linear form in this direction.
Alternatively, the step of judgement characteristic pattern whether there is linear form in this direction are as follows:
Whether the difference of the point probability value of judgement in this direction is less than second threshold；
If being less than, judge whether the sum of point probability value in this direction is greater than first threshold；
If more than then determining that the characteristic pattern has linear form in this direction.
Correspondingly, the present invention also provides a kind of road detection method, the road detection method includes:
Present road image is obtained, and inputs Road detection model and generates present road information, the road detection
Model is the road detection model that the training method supervised training of road detection model described in above-mentioned any one obtains；
Prediction road information is generated according to vehicle movement track and history road information；
Based on Road and the prediction Road of not being blocked, the present road information and the prediction road information are judged
Whether it is associated with；
If association, present road information is corrected according to prediction road information,
Specifically, the prediction road information includes at least prediction Road；
Specifically, the present road information includes at least the Road that is not blocked；
Specifically, expanded Kalman filtration algorithm is used according to prediction road information correction present road information.
Further, before judging whether the present road information is associated with the prediction road information, the road
Line detecting method further include: calculate it is described be not blocked Road between the prediction Road at a distance from.
Further, judge that the step of whether the present road information is associated with the prediction road information includes:
Judge whether the distance is less than preset threshold；
If so, determine that the present road information is associated with the prediction road information,
Specifically, it is reduced using expectation-maximization algorithm and calculates error.
Correspondingly, the present invention also provides a kind of training device of road detection model, the road detection models
Training device include:
Road sample image acquiring unit, for obtaining road sample image.
Extraction unit, for extracting the feature of the road sample image and the further feature of the characteristic pattern.
Specifically, the extraction unit includes the first extraction unit and the second extraction unit:
First extraction unit, for extracting the feature of the road sample image；
Second extraction unit, for extracting the continuity of the characteristic pattern and the further feature of linear form.
Training unit exercises supervision training for the Road data with mark, optimizes deep learning network losses function,
Network parameter is adjusted, road detection model is obtained.
Further, the training device of the road detection model further include:
First judging unit, for judging that the characteristic pattern whether there is continuity in one direction；
Second judgment unit, for judging the characteristic pattern in regional area with the presence or absence of linear form.
The implementation of the embodiments of the present invention has the following beneficial effects:
Training method, the road detection method and device of road detection model disclosed by the invention, using continuous lines
The spatial position continuity of shape operator extraction characteristic pattern and the linear shape of Road, supervised training obtain depth road detection
Model, multiframe integration technology generate prediction road information according to vehicle movement track and history road information, according to prediction road
The part being blocked in line correction present road line, strengthens the robustness of detection, so that road detection model is more suitable for
It is applied in actual condition.
Detailed description of the invention
It in order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology and advantage, below will be to implementation
Example or attached drawing needed to be used in the description of the prior art are briefly described, it should be apparent that, the accompanying drawings in the following description is only
It is only some embodiments of the present invention, for those of ordinary skill in the art, without creative efforts,
It can also be obtained according to these attached drawings other attached drawings.
Fig. 1 is the flow diagram of LaneNet deep learning method；
Fig. 2 is the flow diagram of SCNN deep learning method；
Fig. 3 is a kind of flow diagram of the training method of road detection model provided in an embodiment of the present invention；
Fig. 4 is a kind of flow diagram of road detection method provided in an embodiment of the present invention；
Fig. 5 is a kind of structural schematic diagram of the training device of road detection model provided in an embodiment of the present invention；
Fig. 6 is a kind of interative computation schematic diagram provided in an embodiment of the present invention；
Fig. 7 is the regional area figure of a characteristic pattern；
Fig. 8 is the schematic diagram of the local linear form in a direction.
Specific embodiment
To make the object, technical solutions and advantages of the present invention clearer, the present invention is made into one below in conjunction with attached drawing
Step ground detailed description.Obviously, described embodiment is only one embodiment of the invention, instead of all the embodiments.Base
Embodiment in the present invention, those of ordinary skill in the art are obtained all without creative labor
Other embodiments shall fall within the protection scope of the present invention.
It may be included at least one implementation of the invention it should be noted that " embodiment " referred to herein refers to
A particular feature, structure, or characteristic.In the description of the present invention, term " first ", " second " are used for description purposes only, and cannot manage
Solution is indication or suggestion relative importance or the quantity for implicitly indicating indicated technical characteristic, define as a result, " first ",
The feature of " second " can be expressed or what is implied includes one or more of the features.Also, term " first ", " second "
It is for distinguishing similar object, rather than the specific sequence of description or sequencing.It should be understood that such use data
It can be interchanged in appropriate circumstances, so that the embodiment of the present invention described herein can be in addition to illustrating herein or describing
Those of other than sequence implement.It is to be appreciated that the indicating positions such as term " on ", "lower", "left", "right" or position are closed
System indicates to be based on the orientation or positional relationship shown in the drawings, being merely for convenience of the description present invention or simplifying description
Or imply specific orientation, it is thus impossible to be interpreted as limitation of the present invention.In addition, term " includes " and " for " and its
The deformation of what form, it is intended that cover it is non-exclusive include, for example, containing a series of step and unit, it is not necessary to be limited to
Be clearly listed which step and unit, but can include be not clearly listed or for method and apparatus here its
His step and unit.
Embodiment
Referring to FIG. 3, the stream of its training method for showing a kind of road detection model provided in an embodiment of the present invention
Journey schematic diagram, present description provides method operating procedures as shown in the flowchart, but based on labor conventional or without creativeness
Dynamic may include more or less method operating procedure.The step of enumerating in embodiment sequence is only numerous step operations
One of sequence, does not represent unique operating procedure, can be according to implementation sequence or attached drawing in actual training process
Shown in step operation sequence execute.Specifically as shown in figure 3, the training method of the road detection model includes:
S110 obtains road sample image；
It should be noted that the road sample image includes at least the Road of mark in this specification embodiment
Data.
S120 extracts the feature of the road sample image by convolutional network, obtains characteristic pattern.
S130 uses the further feature of characteristic pattern described in continuous linear operator extraction；
It should be noted that the further feature includes that the spatial position of Road is continuous in this specification embodiment
The linear form of property and Road.
In this specification embodiment, using characteristic pattern described in continuous linear operator extraction further feature the step of wrap
It includes:
Convolution is iterated to the horizontal and vertical of the characteristic pattern using convolution kernel；
Obtain the point probability value of the characteristic pattern in one direction；
According to described probability value, judges the characteristic pattern in this direction and whether there is linear form；
If it exists, then the further feature of the characteristic pattern is extracted；
If it does not exist, then inhibit the direction that linear form is not present.
Specifically, described horizontal and vertical including four sides from right to left, from left to right, from the top down and from bottom to top
To the convolution kernel can be the form of recurrent neural network, be also possible to other forms, and this specification embodiment is not made specifically
It limits.For example, carrying out convolution with each column of the isomorphic convolution kernel to the characteristic pattern, and protect for from left to right
It deposits as a result, carrying out convolution in next column with same group of weight collective effect.Specifically as shown in fig. 6, wherein x indicates input, o table
Show output, U, V, W are weight, and s is to hide layer state.In this specification embodiment, the convolution kernel construction of four direction is consistent,
The weight in each direction is consistent.
Specifically, the step of judgement characteristic pattern whether there is linear form in this direction are as follows:
Whether the sum of the point probability value of judgement in this direction is greater than first threshold；
If more than then judging whether the difference of point probability value in this direction is less than second threshold；
If being less than, determine that the characteristic pattern has linear form in this direction.
Alternatively, the step of judgement characteristic pattern whether there is linear form in this direction are as follows:
Whether the difference of the point probability value of judgement in this direction is less than second threshold；
If being less than, judge whether the sum of point probability value in this direction is greater than first threshold；
If more than then determining that the characteristic pattern has linear form in this direction.
In this specification embodiment, judge whether the sum of probability value is greater than the difference of first threshold and judgement point probability value
Whether second threshold is less than, and the first threshold is a determining value, and the second threshold is also a determining value, judges four respectively
When direction, the first threshold is in the same size, and the second threshold is in the same size.For example, being illustrated in figure 7 the office of a characteristic pattern
Portion's administrative division map, wherein the height of gray scale indicates that there are the probability of Road, as shown in figure 8, a in figure, b, c, d surround central point,
By taking the direction a as an example, calculate 4,5,6 points and 12, the difference of 13,14 points of probability, if the difference greater than first threshold, and less than the
Two threshold values, then it is assumed that the direction a there are linear form, and so on, calculate separately b, c, the direction d whether there is linear form.It is right
The case where there are multiple directions, choosing the maximum direction of the sum of probability is principal direction, and inhibits other directions.
S140 is exercised supervision training with the Road data of mark, optimizes deep learning network according to the further feature
Loss function adjusts network parameter, obtains road detection model.
Referring to FIG. 4, its flow diagram for showing a kind of road detection method provided in an embodiment of the present invention, this
Specification provides method operating procedure as shown in the flowchart, but based on routine or may include more without creative labor
More or less method operating procedure.The step of enumerating in embodiment sequence is only one in numerous step operation sequences
Kind, unique operating procedure is not represented, it, can be according to implementation sequence or step shown in the drawings in actual training process
Operation order executes.Specifically as shown in figure 4, the road detection method includes:
S210 obtains present road image, and inputs Road detection model and generate present road information；
In this specification embodiment, the road detection model is road detection mould described in above-mentioned any one
The road detection model that the training method supervised training of type obtains.
In this specification book embodiment, the present road information includes at least the Road that is not blocked.
S220 generates prediction road information according to vehicle movement track and history road information；
In this specification embodiment, the prediction road information includes at least prediction Road.
S230 is based on be not blocked Road and prediction Road, judges the present road information and the prediction road
Whether information is associated with；
In this specification embodiment, judge whether the present road information is associated with it with the prediction road information
Before, the road detection method further include: calculate it is described be not blocked Road between the prediction Road at a distance from.
In this specification embodiment, the present road information and the whether associated step of prediction road information are judged
Suddenly include:
Judge whether the distance is less than preset threshold；
If so, determining that the present road information is associated with the prediction road information.
If it is not, then determining that the present road information is not associated with the prediction road information.
Specifically, it is reduced using expectation-maximization algorithm and calculates error.
If S240 is associated with, present road information is corrected according to prediction road information；
In this specification embodiment, Extended Kalman filter is used according to prediction road information correction present road information
Algorithm.
Referring to FIG. 5, the knot of its training device for showing a kind of road detection model provided in an embodiment of the present invention
Structure schematic diagram, present description provides the component units as shown in structural schematic diagram, but based on labor conventional or without creativeness
Dynamic may include more or less component units.The component units enumerated in embodiment are only in numerous component units
One kind not representing unique component units, can be component units shown in the drawings in actual structure.Specific such as Fig. 5
Shown, the training device of the road detection model includes:
Road sample image acquiring unit 310, for obtaining road sample image.
Extraction unit 320, for extracting the feature of the road sample image and the further feature of the characteristic pattern.
In this specification embodiment, the extraction unit includes the first extraction unit and the second extraction unit；
First extraction unit, for extracting the feature of the road sample image；
Second extraction unit, for extracting the continuity of the characteristic pattern and the further feature of linear form.
Training unit 330 exercises supervision training for the Road data with mark, optimizes deep learning network losses letter
Number adjusts network parameter, obtains road detection model.
The training device of the road detection model further include:
First judging unit, for judging that the characteristic pattern whether there is continuity in one direction；
Second judgment unit, for judging the characteristic pattern in regional area with the presence or absence of linear form.
Using a kind of training method of road detection model provided in an embodiment of the present invention, road detection method and dress
It sets, using the continuous spatial position continuity of linear operator extraction characteristic pattern and the linear shape of Road, supervised training is obtained
Depth Road detection model, multiframe integration technology generate prediction road according to vehicle movement track and history road information and believe
Breath strengthens the robustness of detection, so that Road is examined according to the part being blocked in prediction Road correction present road line
Model is surveyed to be more suitable for applying in actual condition.
It should be understood that embodiments of the present invention sequencing is for illustration only, do not represent the advantages or disadvantages of the embodiments,
And above-mentioned this specification specific embodiment is described, other embodiments are within the scope of the appended claims.Some
In the case of, the movement recorded in detail in the claims or step can be executed according to the sequence being different from embodiment and
It can be realized desired result.In addition, process depicted in the drawing, which not necessarily requires, shows particular order or consecutive order
It can realize desired result.In some embodiments, multitasking and parallel processing are also possible or possible
It is advantageous.
All the embodiments in this specification are described in a progressive manner, same and similar portion between each embodiment
Dividing may refer to each other, and what each embodiment stressed is the difference with other embodiments.Especially for device
For embodiment, since it is substantially similar to the method embodiment, so being described relatively simple, related place is implemented referring to method
The part explanation of example.
Those of ordinary skill in the art will appreciate that realizing that all or part of step of above-described embodiment can be by hard
Part is completed, and relevant hardware can also be instructed to complete by program, and described program can store in a kind of computer-readable
In medium.
The above is a preferred embodiment of the present invention, it is noted that for those skilled in the art
For, various improvements and modifications may be made without departing from the principle of the present invention, these improvements and modifications are also considered as
Protection scope of the present invention.
Claims (9)
1. a kind of training method of road detection model, which is characterized in that the training method packet of the road detection model
It includes:
Road sample image is obtained, the road sample image includes at least the Road data of mark；
The feature of the road sample image is extracted by convolutional network, obtains characteristic pattern；
Using the further feature of characteristic pattern described in continuous linear operator extraction, wherein the further feature includes the sky of Road
Between position continuity and Road linear form；
According to the further feature, is exercised supervision training with the Road data of mark, optimizes deep learning network losses function,
Network parameter is adjusted, road detection model is obtained.
2. a kind of training method of road detection model according to claim 1, which is characterized in that using continuous linear
The step of further feature of characteristic pattern described in operator extraction includes:
Convolution is iterated to the horizontal and vertical of the characteristic pattern using convolution kernel；
Obtain the point probability value of the characteristic pattern in one direction；
According to described probability value, judges the characteristic pattern in this direction and whether there is linear form；
If it exists, then the further feature of the characteristic pattern is extracted；
If it does not exist, then inhibit the direction that linear form is not present.
3. a kind of training method of road detection model according to claim 2, which is characterized in that described in the judgement
Characteristic pattern whether there is the step of linear form in this direction are as follows:
Whether the sum of the point probability value of judgement in this direction is greater than first threshold；
If more than then judging whether the difference of point probability value in this direction is less than second threshold；
If being less than, determine that the characteristic pattern has linear form in this direction.
4. a kind of road detection method, which is characterized in that the road detection method includes:
Present road image is obtained, and inputs Road detection model and generates present road information, wherein the present road letter
Breath includes at least the Road that is not blocked, and the road detection model is Road described in claim 1-3 any one
The road detection model that the training method supervised training of detection model obtains；
Prediction road information is generated according to vehicle movement track and history road information, wherein the prediction road information is at least
Including predicting Road；
Based on Road and the prediction Road of not being blocked, judge whether are the present road information and the prediction road information
Association；
If association, present road information is corrected according to prediction road information.
5. a kind of road detection method according to claim 4, which is characterized in that judge the present road information with
Before whether the prediction road information is associated with, the road detection method further include:
Calculate it is described be not blocked Road between the prediction Road at a distance from.
6. a kind of road detection method according to claim 5, which is characterized in that judge the present road information with
The step of whether the prediction road information is associated with include:
Judge whether the distance is less than preset threshold；
If so, determining that the present road information is associated with the prediction road information.
7. a kind of training device of road detection model, which is characterized in that the training device packet of the road detection model
It includes:
Road sample image acquiring unit, for obtaining road sample image；
Extraction unit, for extracting the feature of the road sample image and the further feature of the characteristic pattern；
Training unit exercises supervision training for the Road data with mark, optimizes deep learning network losses function, adjustment
Network parameter obtains road detection model.
8. a kind of training device of road detection model according to claim 7, which is characterized in that the extraction unit
Including the first extraction unit and the second extraction unit；
First extraction unit, for extracting the feature of the road sample image；
Second extraction unit, for extracting the further feature of the characteristic pattern.
9. a kind of training device of road detection model according to claim 7, which is characterized in that the Road inspection
Survey the training device of model further include:
First judging unit, for judging that the characteristic pattern whether there is continuity in one direction；
Second judgment unit, for judging the characteristic pattern in regional area with the presence or absence of linear form.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910624027.0A CN110458023B (en) | 2019-07-11 | 2019-07-11 | Training method of road line detection model, and road line detection method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910624027.0A CN110458023B (en) | 2019-07-11 | 2019-07-11 | Training method of road line detection model, and road line detection method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110458023A true CN110458023A (en) | 2019-11-15 |
| CN110458023B CN110458023B (en) | 2022-08-02 |
Family
ID=68482547
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910624027.0A Active CN110458023B (en) | 2019-07-11 | 2019-07-11 | Training method of road line detection model, and road line detection method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110458023B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111291676A (en) * | 2020-02-05 | 2020-06-16 | 清华大学 | Lane line detection method and device based on laser radar point cloud and camera image fusion and chip |
| CN111996883A (en) * | 2020-08-28 | 2020-11-27 | 四川长虹电器股份有限公司 | Method for detecting width of road surface |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105260699A (en) * | 2015-09-10 | 2016-01-20 | 百度在线网络技术（北京）有限公司 | Lane line data processing method and lane line data processing device |
| CN108090456A (en) * | 2017-12-27 | 2018-05-29 | 北京初速度科技有限公司 | A kind of Lane detection method and device |
| US20190130182A1 (en) * | 2017-11-01 | 2019-05-02 | Here Global B.V. | Road modeling from overhead imagery |
-
2019
- 2019-07-11 CN CN201910624027.0A patent/CN110458023B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105260699A (en) * | 2015-09-10 | 2016-01-20 | 百度在线网络技术（北京）有限公司 | Lane line data processing method and lane line data processing device |
| US20190130182A1 (en) * | 2017-11-01 | 2019-05-02 | Here Global B.V. | Road modeling from overhead imagery |
| CN108090456A (en) * | 2017-12-27 | 2018-05-29 | 北京初速度科技有限公司 | A kind of Lane detection method and device |
Non-Patent Citations (2)
| Title |
|---|
| ZE WANG ETAL.: "LaneNet: Real-Time Lane Detection Networks for Autonomous Driving", 《HTTP：ARXIV:1807.01726V1》 * |
| 王丹丹: "基于形态特征的车道线检测和识别技术的研究与实现", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111291676A (en) * | 2020-02-05 | 2020-06-16 | 清华大学 | Lane line detection method and device based on laser radar point cloud and camera image fusion and chip |
| CN111996883A (en) * | 2020-08-28 | 2020-11-27 | 四川长虹电器股份有限公司 | Method for detecting width of road surface |
| CN111996883B (en) * | 2020-08-28 | 2021-10-29 | 四川长虹电器股份有限公司 | Method for detecting width of road surface |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110458023B (en) | 2022-08-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108898047B (en) | Pedestrian detection method and system based on blocking and shielding perception | |
| CN109636771B (en) | Flight target detection method and system based on image processing | |
| WO2017041396A1 (en) | Driving lane data processing method, device, storage medium and apparatus | |
| CN109977790A (en) | A kind of video smoke detection and recognition methods based on transfer learning | |
| CN109118479A (en) | Defects of insulator identification positioning device and method based on capsule network | |
| CN107507170A (en) | A kind of airfield runway crack detection method based on multi-scale image information fusion | |
| CN106228138A (en) | A kind of Road Detection algorithm of integration region and marginal information | |
| CN110889328B (en) | Method, device, electronic equipment and storage medium for detecting road traffic condition | |
| CN107665327B (en) | Lane line detection method and device | |
| CN104794463A (en) | System and method for achieving indoor human body falling detection based on Kinect | |
| CN105809149A (en) | Lane line detection method based on straight lines with maximum length | |
| CN105320927A (en) | Lane line detection method and system | |
| CN107944437B (en) | A kind of Face detection method based on neural network and integral image | |
| CN102750538A (en) | Go competition result analysis method based on image processing technique | |
| CN111626170A (en) | Image identification method for railway slope rockfall invasion limit detection | |
| CN104766071A (en) | Rapid traffic light detection algorithm applied to pilotless automobile | |
| CN106558051A (en) | A kind of improved method for detecting road from single image | |
| CN109961013A (en) | Recognition methods, device, equipment and the computer readable storage medium of lane line | |
| CN112800934B (en) | Behavior recognition method and device for multi-class engineering vehicle | |
| CN108510739A (en) | A kind of road traffic state recognition methods, system and storage medium | |
| CN105243356A (en) | Method of building pedestrian detection model and device and pedestrian detection method | |
| CN110458023A (en) | A kind of training method of road detection model, road detection method and device | |
| CN109740609A (en) | A kind of gauge detection method and device | |
| CN109708658A (en) | A kind of visual odometry method based on convolutional neural networks | |
| CN114494870A (en) | Double-time-phase remote sensing image change detection method, model construction method and device |
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 |