CN108171154A - A kind of method that target detection is combined with lane detection in automatic Pilot - Google Patents
A kind of method that target detection is combined with lane detection in automatic Pilot Download PDFInfo
- Publication number
- CN108171154A CN108171154A CN201711440959.7A CN201711440959A CN108171154A CN 108171154 A CN108171154 A CN 108171154A CN 201711440959 A CN201711440959 A CN 201711440959A CN 108171154 A CN108171154 A CN 108171154A
- Authority
- CN
- China
- Prior art keywords
- detection
- lane
- offset
- automatic pilot
- line
- 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
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course or altitude of land, water, air, or space vehicles, e.g. automatic pilot
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0231—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means
- G05D1/0246—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means using a video camera in combination with image processing means
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Abstract
The method combined the present invention relates to target detection in a kind of automatic Pilot with lane detection more particularly to the vehicle based on CNN and pedestrian detect in real time, lane detection, offset and vehicle lane change information analysis.A kind of method that target detection is combined with lane detection in automatic Pilot, includes the following steps：S1 carries out obtained video image information vehicle and the pedestrian detection in real time based on CNN and analyzes；S2 carries out the detection based on Hough lane lines to obtained video image information, is carried out at the same time offset, the analysis of lane change information and label；Hough lane detections result, offset and lane change information are added in traffic object detection results, and export final result by S3.The present invention real-time target detection and lane detection in automatic Pilot is ensured, promote target and the accuracy of detection and robustness in track.
Description
Technical field
The method that is combined the present invention relates to target detection in a kind of automatic Pilot with lane detection more particularly to based on CNN
Vehicle and pedestrian detect in real time, lane detection, offset and vehicle lane change information analysis.
Background technology
Machine vision, deep learning are at full speed in automatic Pilot field development speed, and machine vision processing is deposited always for many years
It is in the application of convolutional neural networks, these algorithms can extract useful information from the data of the input of sensor.For
The analysis CNN of scene is very efficient, can tell many objects such as automobile, people, animal, road sign, intersection
Point, roadside signs (object), can determine the associated real world object in scene.All these operations are all real-time, as long as passing
After sensor part is provided with, subsequent fusion/decision operation will occur.Either one or all the sensors (LIDAR
(laser acquisition and measure), radar, camera, infrared sensor) input data change, system will at once be made most
Good decision.Such as the scene (far more than the reaction time of driver) braked, automatic vehicle control system can analyze at once away from
From, sensing speed, the reaction time for making brake operation be any manual operation cannot and.
At present, many lane detection algorithms have been proposed in domestic and foreign scholars, are broadly divided into two classes：One kind is based on figure
As the detection method of feature, i.e. character-driven method, be based on road image some features (such as track line color, width and
The features such as edge) by all the points of image labeled as lane line point and non-lane line point, the lane line of this mechanism requirement road
Color is more apparent, and edge is more clear, otherwise can not obtain accurate testing result；Another kind of method is the inspection based on model
Survey method is that the track line model pre-defined is matched according to the feature of extraction, the extraction of lane line is converted into
The computational problem of lane line Model Parameter.
Automatic Pilot technology already become current Hot Contents, but in automatic Pilot sensation target detection with analysis method according to
It is old more to disperse, effectively with reference to multinomial detection content, detection accuracy is promoted with improving robustness, it should to theoretical research and reality
With there is important reference value.
Invention content
For overcome the deficiencies in the prior art, the present invention combines a variety of methods of unmanned middle machine vision and data point
Analysis provides high detection precision in a kind of automatic Pilot, the detection of high robust real-time target and lane detection method.
In order to achieve the goal above, the present invention uses following technical scheme：Provide target detection in a kind of automatic Pilot
The method combined with lane detection, including：Traffic target detection based on CNN includes and creates vehicle and pedestrian's data set；
Improved Hough lane detections, include Hough straight-line detections, and angle keeps track line method with distance restraint, provides offset
Amount and vehicle lane change information.Include the following steps：S1 carries out the vehicle based on convolutional neural networks to obtained video image information
And pedestrian in real time detection and analysis；S2 carries out the detection based on Hough lane lines, simultaneously to obtained video image information
Carry out offset, the analysis of lane change information and label；S3 adds Hough lane detections result, offset and lane change information
Into traffic object detection results, and export final result.
Further, the S1 steps include：S101, specific objective data set are established；S102, vehicle and pedestrian examine in real time
It surveys and analyzes.The S2 steps include：S201, image preprocessing, output image size is M*N, for newly inputting single-frame images
The preparation detected between Hough, including gray processing, binaryzation, Morphological scale-space, canny edge detections and Gaussian Blur；
S202, lane detection, including Hough straight-line detections, the straight line information after detection pretreatment in image, using angle, region
And length constraint, lane line is determined by the way of multi-point fitting, camera lens shakes caused by considering vehicle travel process, using filter
The mode of wave is kept into driveway line；S203, offset and lane change information analysis, by the left and right lane line after the fitting and figure
As the midpoint N/2 of two intersection point of base is compared, regarded as if gained match point abscissa is more than N/2 it is to the left, it is on the contrary then
It regards as to the right, if after larger change occurs for slope (angle) in image coordinate system, regards as lane change.The S3 steps packet
It includes：Lane line and offset and lane change information are added to Overlapping display on object detection results with markup information.
The method that target detection is combined with lane detection in automatic Pilot according to embodiments of the present invention, by specific
Target data set makes, and is detected using the real-time target based on CNN, lowers memory consumption, the accurate of target detection is substantially improved
Rate and detection speed；Using Hough lane detections with deviating and becoming trace analysis, effectively by two visions in automatic Pilot
Detection content combines, and has very high use value and economic interests.
Description of the drawings
With reference to attached drawing, by detailed description below, it can be more clearly understood that the above-mentioned and other feature of the present invention and excellent
Point, wherein：
Fig. 1 is the module map combined according to the target detection of the embodiment of the present invention with track；
Fig. 2 is the Hough lane detections according to the embodiment of the present invention, offset and lane change information analysis module figure；
Fig. 3 is to establish scheme according to the image coordinate system of the embodiment of the present invention；
Fig. 4 is the lane line range constraint figure according to the embodiment of the present invention；
Fig. 5 is according to lane detection of the embodiment of the present invention or so track fitted figure；
Fig. 6 is the offset analysis chart according to the embodiment of the present invention.
Specific embodiment
Referring to the attached drawing of the specific embodiment of the invention, the present invention is described in more detail.However, the present invention can be with
Many different forms are realized, and should not be construed as being limited by the embodiment herein proposed.On the contrary, propose that these embodiments are
In order to reach abundant and complete disclosure, and those skilled in the art is made to understand the scope of the present invention completely.
Target detection and lane detection combined method, packet in automatic Pilot according to embodiments of the present invention is now described in detail
It includes：Traffic target detection based on CNN includes and creates vehicle and pedestrian's data set；Improved Hough lane detections,
Hough straight-line detections are included, angle keeps track line method with distance restraint, provides offset and vehicle lane change information.Including
Following steps：S1 carries out obtained video image information vehicle and the pedestrian detection in real time based on convolutional neural networks and divides
Analysis；S2 carries out the detection based on Hough lane lines to obtained video image information, is carried out at the same time offset, lane change information
Analysis and label；Hough lane detections result, offset and lane change information are added to traffic target detection knot by S3
In fruit, and export final result.The S1 steps include：S101, specific objective data set are established；S102, vehicle and pedestrian are real
When detection and analysis.The S2 steps include：S201, image preprocessing, output image size is M*N, for newly inputting single frames
The preparation detected between image Hough, including gray processing, binaryzation, Morphological scale-space, canny edge detections and Gaussian Blur；
S202, lane detection, including Hough straight-line detections, the straight line information after detection pretreatment in image, using angle, region
And length constraint, lane line is determined by the way of multi-point fitting, camera lens shakes caused by considering vehicle travel process, using filter
The mode of wave is kept into driveway line；S203, offset and lane change information analysis, by the left and right lane line after the fitting and figure
As the midpoint N/2 of two intersection point of base is compared, regarded as if gained match point abscissa is more than N/2 it is to the left, it is on the contrary then
It regards as to the right, if after larger change occurs for slope (angle) in image coordinate system, regards as lane change.The S3 steps packet
It includes：Lane line and offset and lane change information are added to Overlapping display on object detection results with markup information.
The present invention is altogether there are two module composition, and one is that CNN vehicles and pedestrian target to detect in real time, and two be Hough lane lines
Detection and offset and lane change information analysis.
Procedure structure is realized as shown in Figure 1, input video single frames is to CNN target detections and Hough lane detections, by vehicle
Diatom and offset are marked with lane change information result onto CNN testing results, export final detection result.
Hough lane lines and offset and lane change information analysis structure by single-frame images as shown in Fig. 2, pre-processed, Hough
Straight-line detection, straight line angle length constraint, multi-point fitting and offset are with becoming trace analysis.
Fig. 3 establishes scheme for image coordinate system, and region characterizing portion is mainly that attached drawing 4 shows content.
Lane line is constrained to be realized with fitting, before and after fitting specifically as shown in Figure 5, the slope line specially in threshold range
Section, line segment length are then marked its two-end-point more than certain value, if it exceeds 2 points, then using multi-point fitting, otherwise adopt
It is fitted with 2 points, and obtains the straight line after fitting and intersection point above and below limited area, realize the fitting of lane line.
Offset analysis principle, can be with reference to attached drawing 6, specially：Lane line and image base intersection point A, B after being fitted is obtained to sit
Mark, intersection point is respectively the straight line after fitting up and down with limited area：A2(XA2,YA2)、B2(XB2,YB2)、A1(XA1,YA1)、B1
(XB1,YB1).And count base intersection point A1 (XA1,YA1) and B1 (XB1,YB1) midpoint MABCoordinate, base central point M coordinate informations.
According to the proportionate relationship of actual range and image distance, according to M and MABInterval calculation offset distance.
Lane change analysis principle, can be with reference to attached drawing 6, specially：When offset reaches certain threshold value, A1A2 or B1B2 slope values
Restriction range is exceeded, has been determined as lane change, deviation left side is more, is left lane change, is otherwise lane change to the right.Wherein, A1A2 and B1B2
The slope of lane line is respectively：
Target detection and lane detection combined method in automatic Pilot according to embodiments of the present invention, by specific objective
Data set makes, and is detected using based on CNN real-time targets, reduces memory loss, promotes detection speed；Using Hough lane lines
Two vision-based detection contents in automatic Pilot are effectively combined with deviating and becoming trace analysis, have very high make by detection
With value and economic interests.
The preferred embodiment of the present invention described in detail above.It should be appreciated that those of ordinary skill in the art without
Creative work is needed according to the present invention can to conceive and makes many modifications and variations.All technician in the art according to
The design of the present invention passes through the available technical side of logical analysis, reasoning, or a limited experiment on the basis of existing technology
Case, all should be in the protection domain being defined in the patent claims.
Claims (4)
1. a kind of method that target detection is combined with lane detection in automatic Pilot, which is characterized in that include the following steps：S1,
Vehicle and the pedestrian detection in real time based on CNN are carried out to obtained video image information and is analyzed；S2, to obtained video image
Information carries out the detection based on Hough lane lines, is carried out at the same time offset, the analysis of lane change information and label；S3, by Hough
Lane detection result, offset and lane change information are added in traffic object detection results, and export final result.
2. the method that target detection is combined with lane detection in automatic Pilot according to claim 1, which is characterized in that institute
S1 steps are stated to include：S101, specific objective data set are established；S102, vehicle and pedestrian detection in real time and analysis.
3. the method that target detection is combined with lane detection in automatic Pilot according to claim 1, which is characterized in that institute
S2 steps are stated to include：S201, image preprocessing, output image size is M*N, is examined for newly inputting between single-frame images Hough
The preparation of survey, including gray processing, binaryzation, Morphological scale-space, canny edge detections and Gaussian Blur；S202, lane line inspection
It surveys, including Hough straight-line detections, the straight line information after detection pretreatment in image using angle, region and length constraint, is adopted
Lane line is determined with the mode of multi-point fitting, is kept by the way of filtering into driveway line；S203, offset and lane change information
Analysis, the midpoint N/2 of the left and right lane line after the fitting and two intersection point of image base is compared, if gained match point
Abscissa then regarded as more than N/2 it is to the left, it is on the contrary then regard as it is to the right, if slope (angle) occurs larger to change in image coordinate system
After change, then lane change is regarded as.
4. the method that target detection is combined with lane detection in automatic Pilot according to claim 1, which is characterized in that institute
S3 steps are stated to include：Lane line and offset and lane change information are added to Overlapping display on object detection results with markup information.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711440959.7A CN108171154A (en) | 2017-12-27 | 2017-12-27 | A kind of method that target detection is combined with lane detection in automatic Pilot |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711440959.7A CN108171154A (en) | 2017-12-27 | 2017-12-27 | A kind of method that target detection is combined with lane detection in automatic Pilot |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108171154A true CN108171154A (en) | 2018-06-15 |
Family
ID=62522093
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711440959.7A Pending CN108171154A (en) | 2017-12-27 | 2017-12-27 | A kind of method that target detection is combined with lane detection in automatic Pilot |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108171154A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109407679A (en) * | 2018-12-28 | 2019-03-01 | 百度在线网络技术（北京）有限公司 | Method and apparatus for controlling pilotless automobile |
| CN110525436A (en) * | 2019-08-27 | 2019-12-03 | 中国第一汽车股份有限公司 | Vehicle lane-changing control method, device, vehicle and storage medium |
| CN110530863A (en) * | 2019-08-27 | 2019-12-03 | 南京末梢信息技术有限公司 | A kind of automotive hub mixes package detection device and method |
| CN110967025A (en) * | 2018-09-30 | 2020-04-07 | 长城汽车股份有限公司 | Lane line screening method and system |
| CN111127651A (en) * | 2020-03-31 | 2020-05-08 | 江苏广宇科技产业发展有限公司 | Automatic driving test development method and device based on high-precision visualization technology |
| CN111383464A (en) * | 2018-12-28 | 2020-07-07 | 沈阳美行科技有限公司 | Vehicle lane change recognition method and device, electronic equipment and medium |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105632186A (en) * | 2016-03-11 | 2016-06-01 | 博康智能信息技术有限公司 | Method and device for detecting vehicle queue jumping behavior |
| CN105966398A (en) * | 2016-06-21 | 2016-09-28 | 广州鹰瞰信息科技有限公司 | Method and device for early warning lane departure of vehicle |
| CN106056100A (en) * | 2016-06-28 | 2016-10-26 | 重庆邮电大学 | Vehicle auxiliary positioning method based on lane detection and object tracking |
| CN106652468A (en) * | 2016-12-09 | 2017-05-10 | 武汉极目智能技术有限公司 | Device and method for detection of violation of front vehicle and early warning of violation of vehicle on road |
| CN107133600A (en) * | 2017-05-11 | 2017-09-05 | 南宁市正祥科技有限公司 | A kind of real-time lane line detection method based on intra-frame trunk |
| CN107481526A (en) * | 2017-09-07 | 2017-12-15 | 公安部第三研究所 | System and method for drive a vehicle lane change detection record and lane change violating the regulations report control |
| CN107516078A (en) * | 2017-08-18 | 2017-12-26 | 电子科技大学 | A kind of lane detection fixing method |
-
2017
- 2017-12-27 CN CN201711440959.7A patent/CN108171154A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105632186A (en) * | 2016-03-11 | 2016-06-01 | 博康智能信息技术有限公司 | Method and device for detecting vehicle queue jumping behavior |
| CN105966398A (en) * | 2016-06-21 | 2016-09-28 | 广州鹰瞰信息科技有限公司 | Method and device for early warning lane departure of vehicle |
| CN106056100A (en) * | 2016-06-28 | 2016-10-26 | 重庆邮电大学 | Vehicle auxiliary positioning method based on lane detection and object tracking |
| CN106652468A (en) * | 2016-12-09 | 2017-05-10 | 武汉极目智能技术有限公司 | Device and method for detection of violation of front vehicle and early warning of violation of vehicle on road |
| CN107133600A (en) * | 2017-05-11 | 2017-09-05 | 南宁市正祥科技有限公司 | A kind of real-time lane line detection method based on intra-frame trunk |
| CN107516078A (en) * | 2017-08-18 | 2017-12-26 | 电子科技大学 | A kind of lane detection fixing method |
| CN107481526A (en) * | 2017-09-07 | 2017-12-15 | 公安部第三研究所 | System and method for drive a vehicle lane change detection record and lane change violating the regulations report control |
Non-Patent Citations (2)
| Title |
|---|
| 吕亚运: "基于机器视觉的车道线偏移检测与预警系统设计与实现", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 鱼兆伟: "基于动态感兴趣区域的光照无关车道线检测算法", 《计算机工程》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110967025A (en) * | 2018-09-30 | 2020-04-07 | 长城汽车股份有限公司 | Lane line screening method and system |
| CN109407679A (en) * | 2018-12-28 | 2019-03-01 | 百度在线网络技术（北京）有限公司 | Method and apparatus for controlling pilotless automobile |
| CN111383464A (en) * | 2018-12-28 | 2020-07-07 | 沈阳美行科技有限公司 | Vehicle lane change recognition method and device, electronic equipment and medium |
| CN111383464B (en) * | 2018-12-28 | 2022-11-18 | 沈阳美行科技股份有限公司 | Vehicle lane change recognition method and device, electronic equipment and medium |
| CN110525436A (en) * | 2019-08-27 | 2019-12-03 | 中国第一汽车股份有限公司 | Vehicle lane-changing control method, device, vehicle and storage medium |
| CN110530863A (en) * | 2019-08-27 | 2019-12-03 | 南京末梢信息技术有限公司 | A kind of automotive hub mixes package detection device and method |
| CN110530863B (en) * | 2019-08-27 | 2022-04-05 | 南京末梢信息技术有限公司 | Automobile hub mixed package detection device and method |
| CN111127651A (en) * | 2020-03-31 | 2020-05-08 | 江苏广宇科技产业发展有限公司 | Automatic driving test development method and device based on high-precision visualization technology |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108171154A (en) | A kind of method that target detection is combined with lane detection in automatic Pilot | |
| CN105261020B (en) | A kind of express lane line detecting method | |
| CN107576960B (en) | Target detection method and system for visual radar space-time information fusion | |
| CN105260699B (en) | A kind of processing method and processing device of lane line data | |
| CN104246821B (en) | Three-dimensional body detection device and three-dimensional body detection method | |
| CN105336169B (en) | A kind of method and system that traffic congestion is judged based on video | |
| CN107025432B (en) | A kind of efficient lane detection tracking and system | |
| CN102682292B (en) | Method based on monocular vision for detecting and roughly positioning edge of road | |
| Nieto et al. | Road environment modeling using robust perspective analysis and recursive Bayesian segmentation | |
| CN103797529B (en) | Three-dimensional body detects device | |
| CN110738121A (en) | front vehicle detection method and detection system | |
| CN108280450A (en) | A kind of express highway pavement detection method based on lane line | |
| CN110379168B (en) | Traffic vehicle information acquisition method based on Mask R-CNN | |
| CN109460709A (en) | The method of RTG dysopia analyte detection based on the fusion of RGB and D information | |
| CN107590470B (en) | Lane line detection method and device | |
| CN108596081A (en) | A kind of traffic detection method merged based on radar and video camera | |
| CN108171695A (en) | A kind of express highway pavement detection method based on image procossing | |
| CN103366156A (en) | Road structure detection and tracking | |
| CN105551264A (en) | Speed detection method based on license plate characteristic matching | |
| CN109446919A (en) | A kind of vision lane keeping method based on end-to-end study | |
| Mulyanto et al. | Real-time human detection and tracking using two sequential frames for advanced driver assistance system | |
| CN102768726A (en) | Pedestrian detection method for preventing pedestrian collision | |
| CN111259796A (en) | Lane line detection method based on image geometric features | |
| CN112793567A (en) | Driving assistance method and system based on road condition detection | |
| CN108764338A (en) | A kind of pedestrian tracking algorithm applied to video analysis |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| 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 | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20180615 |