CN108764075A - The method of container truck location under gantry crane - Google Patents
The method of container truck location under gantry crane Download PDFInfo
- Publication number
- CN108764075A CN108764075A CN201810459059.5A CN201810459059A CN108764075A CN 108764075 A CN108764075 A CN 108764075A CN 201810459059 A CN201810459059 A CN 201810459059A CN 108764075 A CN108764075 A CN 108764075A
- Authority
- CN
- China
- Prior art keywords
- lane line
- line
- lane
- vehicle
- gantry crane
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Control And Safety Of Cranes (AREA)
Abstract
The present invention relates to a kind of methods of container truck location under gantry crane, it is characterized in that：The lane line under gantry crane is detected by deep learning, the positioning of vehicle is realized, is as follows：By deep learning network, the lane line region in image is identified and is divided, lane detection model is established in the region of segmentation；Lane line accurately detects：By row scanning, depth-first search, fitting a straight line；Vehicle location：Lateral direction of car positioning, longitudinal direction of car positioning.Advantageous effect：The present invention can help unpiloted container truck to be positioned in gantry crane, lay the foundation for the unmanned of port and pier.
Description
Technical field
The invention belongs to a kind of methods of container truck location under intelligent transportation field more particularly to gantry crane.
Background technology
Currently used vehicle positioning technology is based on the GPS device on vehicle.However, due to the screening of bridge crane in gantry crane
Gear, signal strength is inadequate, and GPS device cannot achieve positioning function, therefore can not based on the localization method that traditional GPS device carries out
Vehicle under this environment is positioned.
Lane line in gantry crane is different from the lane line on common road.For example, the track line color in gantry crane is yellow,
And width is different from normal road, therefore, also can not work normally based on the lane line localization method on normal road.
Invention content
It is an object of the invention to overcome the defect of prior art, and provide a kind of method of container truck location under gantry crane,
The lane line under gantry crane is detected by deep learning, realizes the positioning of vehicle.
The present invention to achieve the above object, is achieved through the following technical solutions, container truck location under a kind of gantry crane
Method, it is characterized in that：The lane line under gantry crane is detected by deep learning, realizes the positioning of vehicle, specific steps are such as
Under：
1) lane detection model training
1. installing camera in the front of experiment vehicle, got off back the true operation feelings of driving vehicle with simulating gantry crane
Condition acquires the track line image under gantry crane, i.e. Vehicular video image data set；
2. image labeling：Manual mark is carried out to the lane line in the image collected, to distinguish lane line in image
Region and non-lane line region；
3. deep learning network carries out off-line training：Using annotation results 2. as the input data of deep learning network,
Lane line region in image is identified and is divided, for training lane detection model deep learning network, is being divided
Region establish lane detection model, the region detected, be partitioned into image where lane line in real time；
4. lane detection model：After the segmentation of lane detection model, the probability value of lane line is indicated using pixel, it will be general
Rate value zooms to 0 to 255 section, obtains gray-scale map, and the probability of brighter Regional Representative's lane line is bigger wherein in gray-scale map, after
Continuous lane detection process carries out on this gray-scale map；
2) lane line accurately detects
For the lane detection model of foundation,
1. by row scanning：The gray-scale map at above-mentioned zone center is set as the center of lane line, to continuous per a line in gray-scale map
Region, by row scanning, seek " center of gravity " of each piece of continuum, be used in combination the center of gravity replace the region, row scan after the completion of,
Binary picture is obtained,
2. depth-first search forms the set of same lane line：After the completion of being scanned by row, finds out and belong to this track
The pixel of line carries out fitting a straight line using these pixels, and the adjacent pixels for the pixel that binaryzation intermediate value is 1 are formed a collection
It closes, considers each set when fitting a straight line successively；
3. fitting a straight line：Each set carries out fitting a straight line using RANSAC algorithms to the point set, and the straight line of fitting is
For potential lane line；
3) vehicle location：Located lateral including vehicle, the i.e. distance of vehicle distances lane line；The longitudinal register of vehicle,
The distance that i.e. vehicle is travelled relative to initial position；
Lateral direction of car positions：3. the anti-perspective mapping transformation carried out centered on vehicle location, walks in conjunction in step 2) the
It detects potential lane line, calculates vehicle to the distance of lane line, the i.e. distance of image center to lane line, complete vehicle
Located lateral；
Longitudinal direction of car positions：It is integrated using wheel speed meter, the distance that vehicle is within a specified time travelled, completes vehicle
Longitudinal register.
Carry out rejecting the lane line of flase drop after the depth-first search of the step of lane line accurately detects 2., it is specific to walk
Suddenly it is,
1. carrying out anti-perspective mapping transformation to the straight line detected, using most rectilinear direction of counting as benchmark, reject
The straight line of given threshold value is differed by more than with the straight line angle；
2. calculate the distance between straight line two-by-two, when the distance of two straight lines is less than given threshold value, delete points compared with
That few straight line deletes the lane line that the accurate detecting step flase drop of lane line comes out.
Advantageous effect：Compared with prior art, due to illumination condition bad (such as illumination is too strong or excessively dark), the shadow of shade
It rings, the image taken by video camera is unsatisfactory, but there are certain noises.The presence of these noises increases inspection
The difficulty of measuring car diatom.The present invention is detected the lane line under gantry crane by deep learning, realizes the positioning of vehicle, can be with
It helps unpiloted container truck to be positioned in gantry crane, lays the foundation for the unmanned of port and pier.
Description of the drawings
Fig. 1 is the lane detection model training flow chart of the present invention；
Fig. 2 is the flow chart of lane detection of the present invention.
Specific implementation mode
Below in conjunction with preferred embodiment, to the specific implementation mode that provides according to the present invention, details are as follows：
Attached drawing is referred to, present embodiment discloses a kind of methods of container truck location under gantry crane, it is characterized in that：By depth
Degree study is detected the lane line under gantry crane, realizes the positioning of vehicle, is as follows：
1) lane detection model training
1. installing camera in the front of experiment vehicle, got off back the true operation feelings of driving vehicle with simulating gantry crane
Condition acquires the track line image under gantry crane, i.e. Vehicular video image data set；
2. image labeling：Manual mark is carried out to the lane line in the image collected, to distinguish lane line in image
Region and non-lane line region；
3. deep learning network carries out off-line training：Using annotation results 2. as the input data of deep learning network,
Lane line region in image is identified and is divided, for training lane detection model deep learning network, is being divided
Region establish lane detection model, the region detected, be partitioned into image where lane line in real time；
4. lane detection model：After the segmentation of lane detection model lane line detection model, track is indicated using pixel
Probability value is zoomed to 0 to 255 section, obtains gray-scale map, brighter Regional Representative track wherein in gray-scale map by the probability value of line
The probability of line is bigger, and subsequent lane detection process carries out on this gray-scale map；
2) lane line accurately detects
For the lane detection model of foundation,
1. by row scanning：The gray-scale map at above-mentioned zone center is set as the center of lane line, to continuous per a line in gray-scale map
Region, by row scanning, seek " center of gravity " of each piece of continuum, be used in combination the center of gravity replace the region, row scan after the completion of,
Binary picture is obtained,
2. depth-first search forms the set of same lane line：After the completion of being scanned by row, finds out and belong to this track
The pixel of line carries out fitting a straight line using these pixels, and the adjacent pixels for the pixel that binaryzation intermediate value is 1 are formed a collection
It closes, considers each set when fitting a straight line successively；
3. fitting a straight line：Each set carries out fitting a straight line using RANSAC algorithms to the point set, and the straight line of fitting is
For potential lane line；
3) vehicle location：Located lateral including vehicle, the i.e. distance of vehicle distances lane line；The longitudinal register of vehicle,
The distance that i.e. vehicle is travelled relative to initial position；
Lateral direction of car positions：3. the anti-perspective mapping transformation carried out centered on vehicle location, walks in conjunction in step 2) the
It detects potential lane line, calculates vehicle to the distance of lane line, the i.e. distance of image center to lane line, complete vehicle
Located lateral；
Longitudinal direction of car positions：It is integrated using wheel speed meter, the distance that vehicle is within a specified time travelled, completes vehicle
Longitudinal register.
Carried out after the depth-first search of the step of lane line accurately detects 2. reject flase drop lane line, in order into
One step removes the influence of noise, can delete set of the points less than a given value.It comprises the concrete steps that,
1. carrying out anti-perspective mapping transformation to the straight line detected, using most rectilinear direction of counting as benchmark, reject
The straight line of given threshold value is differed by more than with the straight line angle；
2. calculate the distance between straight line two-by-two, when the distance of two straight lines is less than given threshold value, delete points compared with
That few straight line deletes the lane line that the accurate detecting step flase drop of lane line comes out.
Operation principle
The present invention can be identified the lane line region in image, divide, then in the region of segmentation into runway
Line detects.Due to area reduction to be detected after segmentation, the influence of noise is also suppressed, to improve lane detection
Precision.
The image of lane line is acquired by installing a camera on truck and truck being allowed back and forth to be travelled in gantry crane
Data；Secondly, manual mark is carried out to the image collected, distinguishes lane line and non-lane line region；Then, the present invention by
Deep learning network is trained the data of mark, to generate lane detection model.It is worth noting that, lane line is examined
The training for surveying model needs to expend the long time；After training, the model of generation can carry out the image of input
Segmentation in real time, obtains lane line region.The size and original input picture in the region are in the same size, wherein each pixel generation
Table its whether belong to the probability of lane line.Subsequent lane detection work is only limited in this region split.
Wherein step 2) the lane line accurately depth-first search of detection 2.：By row scan after the completion of, we obtain only
It is the pixel for belonging to lane line center.In order to fit a lane line, we must be the picture for belonging to this lane line first
Element is all found out, and then carries out fitting a straight line using these pixels.In order to find out the pixel for belonging to certain lane line, Wo Menke
To utilize following priori:(a) two different lane lines are all separated by a distance；(b) any one lane line be all
There is certain length.Therefore, the pixel for belonging to same lane line should be adjacent in binary image obtained in the previous step
Either the distance between the pixel connect is less than a threshold value.In this regard, we can by the depth-first search in graph theory,
The adjacent pixels for the pixel that binaryzation intermediate value is 1 are all found out, and a set is formed.Point in each set is to belong to
Same lane line.When to carry out fitting a straight line, it is only necessary to consider that each set can successively.
The above-mentioned detailed description carried out to the method for container truck location under a kind of gantry crane with reference to embodiment, is explanation
Property without being restrictive, several embodiments can be enumerated according to limited range, therefore of the invention overall not departing from
Change and modification under design, should belong within protection scope of the present invention.
Claims (2)
1. a kind of method of container truck location under gantry crane, it is characterized in that：By deep learning to the lane line under gantry crane into
Row detection, realizes the positioning of vehicle, is as follows：
1) lane detection model training
1. installing camera in the front of experiment vehicle, is got off back the true handling situations of driving vehicle with simulating gantry crane, adopted
Collect the track line image under gantry crane, i.e. Vehicular video image data set；
2. image labeling：Manual mark is carried out to the lane line in the image collected, to distinguish lane line region in image
With non-lane line region；
3. deep learning network carries out off-line training：Using annotation results 2. as the input data of deep learning network, to figure
Lane line region as in is identified and divides, for training lane detection model deep learning network, in the area of segmentation
Lane detection model is established in domain, the region detected, be partitioned into image where lane line in real time；
4. lane detection model：After the segmentation of lane detection model, the probability value of lane line is indicated using pixel, by probability value
0 to 255 section is zoomed to, obtains gray-scale map, the probability of brighter Regional Representative's lane line is bigger wherein in gray-scale map, subsequent
Lane detection process carries out on this gray-scale map；
2) lane line accurately detects
For the lane detection model of foundation,
1. by row scanning：The gray-scale map at above-mentioned zone center is set as the center of lane line, to every continuous area of a line in gray-scale map
" center of gravity " of each piece of continuum is sought in domain by row scanning, is used in combination the center of gravity that the region is replaced to be obtained after the completion of row scanning
Binary picture,
2. depth-first search forms the set of same lane line：After the completion of being scanned by row, finds out and belong to this lane line
Pixel carries out fitting a straight line using these pixels, the adjacent pixels for the pixel that binaryzation intermediate value is 1 is formed a set, directly
Each set is considered when line is fitted successively；
3. fitting a straight line：Each set carries out fitting a straight line using RANSAC algorithms to the point set, and the straight line of fitting is latent
Lane line；
3) vehicle location：Located lateral including vehicle, the i.e. distance of vehicle distances lane line；The longitudinal register of vehicle, i.e. vehicle
Relative to initial position traveling distance；
Lateral direction of car positions：The anti-perspective mapping transformation carried out centered on vehicle location, in conjunction with 3. step detection in step 2)
Go out potential lane line, calculates vehicle to the distance of lane line, the i.e. distance of image center to lane line, complete vehicle
Located lateral；
Longitudinal direction of car positions：It is integrated using wheel speed meter, the distance that vehicle is within a specified time travelled, completes the vertical of vehicle
To positioning.
2. the method for container truck location under gantry crane according to claim 1, it is characterized in that：The lane line is accurately examined
The lane line for carrying out rejecting flase drop after the depth-first search of the step of survey 2., comprises the concrete steps that,
1. carrying out anti-perspective mapping transformation to the straight line detected, using most rectilinear direction of counting as benchmark, rejects and be somebody's turn to do
Straight line angle differs by more than the straight line of given threshold value；
2. calculating the distance between straight line two-by-two, when the distance of two straight lines is less than given threshold value, it is less to delete points
That straight line deletes the lane line that the accurate detecting step flase drop of lane line comes out.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810459059.5A CN108764075A (en) | 2018-05-15 | 2018-05-15 | The method of container truck location under gantry crane |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810459059.5A CN108764075A (en) | 2018-05-15 | 2018-05-15 | The method of container truck location under gantry crane |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108764075A true CN108764075A (en) | 2018-11-06 |
Family
ID=64006768
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810459059.5A Pending CN108764075A (en) | 2018-05-15 | 2018-05-15 | The method of container truck location under gantry crane |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108764075A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109489680A (en) * | 2018-12-29 | 2019-03-19 | 百度在线网络技术（北京）有限公司 | A kind of the reference locus line generation method and mobile unit of screw lane car |
| CN109961013A (en) * | 2019-02-21 | 2019-07-02 | 杭州飞步科技有限公司 | Recognition methods, device, equipment and the computer readable storage medium of lane line |
| CN110008851A (en) * | 2019-03-15 | 2019-07-12 | 深兰科技（上海）有限公司 | A kind of method and apparatus of lane detection |
| CN112232285A (en) * | 2020-11-05 | 2021-01-15 | 浙江点辰航空科技有限公司 | Unmanned aerial vehicle system that highway emergency driveway was patrolled and examined |
| CN112415548A (en) * | 2020-11-09 | 2021-02-26 | 北京斯年智驾科技有限公司 | Unmanned card-collecting positioning method, device and system, electronic device and storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102398598A (en) * | 2010-07-20 | 2012-04-04 | 通用汽车环球科技运作有限责任公司 | Lane fusion system using forward-view and rear-view cameras |
| CN103940434A (en) * | 2014-04-01 | 2014-07-23 | 西安交通大学 | Real-time lane line detecting system based on monocular vision and inertial navigation unit |
| CN105631880A (en) * | 2015-12-31 | 2016-06-01 | 百度在线网络技术（北京）有限公司 | Lane line segmentation method and apparatus |
| CN106500705A (en) * | 2016-10-20 | 2017-03-15 | 上海智驾电子科技有限公司 | Alignment system and localization method based on lane detection |
| CN107339996A (en) * | 2017-06-30 | 2017-11-10 | 百度在线网络技术（北京）有限公司 | Vehicle method for self-locating, device, equipment and storage medium |
| US20180129887A1 (en) * | 2016-11-07 | 2018-05-10 | Samsung Electronics Co., Ltd. | Method and apparatus for indicating lane |
-
2018
- 2018-05-15 CN CN201810459059.5A patent/CN108764075A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102398598A (en) * | 2010-07-20 | 2012-04-04 | 通用汽车环球科技运作有限责任公司 | Lane fusion system using forward-view and rear-view cameras |
| CN103940434A (en) * | 2014-04-01 | 2014-07-23 | 西安交通大学 | Real-time lane line detecting system based on monocular vision and inertial navigation unit |
| CN105631880A (en) * | 2015-12-31 | 2016-06-01 | 百度在线网络技术（北京）有限公司 | Lane line segmentation method and apparatus |
| CN106500705A (en) * | 2016-10-20 | 2017-03-15 | 上海智驾电子科技有限公司 | Alignment system and localization method based on lane detection |
| US20180129887A1 (en) * | 2016-11-07 | 2018-05-10 | Samsung Electronics Co., Ltd. | Method and apparatus for indicating lane |
| CN107339996A (en) * | 2017-06-30 | 2017-11-10 | 百度在线网络技术（北京）有限公司 | Vehicle method for self-locating, device, equipment and storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 吴平: "基于视觉传感器的结构化道路车道线识别算法研究", 《中国优秀硕士学位论文全文数据库-工程科技Ⅱ辑》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109489680A (en) * | 2018-12-29 | 2019-03-19 | 百度在线网络技术（北京）有限公司 | A kind of the reference locus line generation method and mobile unit of screw lane car |
| CN109961013A (en) * | 2019-02-21 | 2019-07-02 | 杭州飞步科技有限公司 | Recognition methods, device, equipment and the computer readable storage medium of lane line |
| CN110008851A (en) * | 2019-03-15 | 2019-07-12 | 深兰科技（上海）有限公司 | A kind of method and apparatus of lane detection |
| CN110008851B (en) * | 2019-03-15 | 2021-11-19 | 深兰科技（上海）有限公司 | Method and equipment for detecting lane line |
| CN112232285A (en) * | 2020-11-05 | 2021-01-15 | 浙江点辰航空科技有限公司 | Unmanned aerial vehicle system that highway emergency driveway was patrolled and examined |
| CN112415548A (en) * | 2020-11-09 | 2021-02-26 | 北京斯年智驾科技有限公司 | Unmanned card-collecting positioning method, device and system, electronic device and storage medium |
| CN112415548B (en) * | 2020-11-09 | 2023-09-29 | 北京斯年智驾科技有限公司 | Positioning method, device and system of unmanned integrated card, electronic device and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108764075A (en) | The method of container truck location under gantry crane | |
| CN105488454B (en) | Front vehicles detection and ranging based on monocular vision | |
| CN106997688B (en) | Parking lot parking space detection method based on multi-sensor information fusion | |
| CN100403332C (en) | Vehicle lane Robust identifying method for lane deviation warning | |
| Yu et al. | Automated detection of urban road manhole covers using mobile laser scanning data | |
| CN103383733B (en) | A kind of track based on half machine learning video detecting method | |
| JP2917661B2 (en) | Traffic flow measurement processing method and device | |
| EP1975565A2 (en) | Feature information collecting apparatus and feature information collecting method | |
| CN104282020A (en) | Vehicle speed detection method based on target motion track | |
| CN104392212A (en) | Method for detecting road information and identifying forward vehicles based on vision | |
| CN108280450A (en) | A kind of express highway pavement detection method based on lane line | |
| JP4032727B2 (en) | Lane boundary detection device | |
| CN109635737B (en) | Auxiliary vehicle navigation positioning method based on road marking line visual identification | |
| CN110197173B (en) | Road edge detection method based on binocular vision | |
| CN103324913A (en) | Pedestrian event detection method based on shape features and trajectory analysis | |
| CN108171695A (en) | A kind of express highway pavement detection method based on image procossing | |
| CN112070756B (en) | Three-dimensional road surface disease measuring method based on unmanned aerial vehicle oblique photography | |
| CN107798293A (en) | A kind of crack on road detection means | |
| CN110379168A (en) | A kind of vehicular traffic information acquisition method based on Mask R-CNN | |
| CN108171154A (en) | A kind of method that target detection is combined with lane detection in automatic Pilot | |
| Wu et al. | Adjacent lane detection and lateral vehicle distance measurement using vision-based neuro-fuzzy approaches | |
| CN107067752A (en) | Automobile speedestimate system and method based on unmanned plane image | |
| CN108573244B (en) | Vehicle detection method, device and system | |
| CN111047879A (en) | Vehicle overspeed detection method | |
| Janda et al. | Road boundary detection for run-off road prevention based on the fusion of video and radar |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20181106 |
|
| WD01 | Invention patent application deemed withdrawn after publication |