CN109614966B - It is a kind of based on information fusion Lidar sensor efficient road surface and curb detection method - Google Patents
It is a kind of based on information fusion Lidar sensor efficient road surface and curb detection method Download PDFInfo
- Publication number
- CN109614966B CN109614966B CN201910169999.5A CN201910169999A CN109614966B CN 109614966 B CN109614966 B CN 109614966B CN 201910169999 A CN201910169999 A CN 201910169999A CN 109614966 B CN109614966 B CN 109614966B
- Authority
- CN
- China
- Prior art keywords
- road surface
- point
- curb
- cloud
- point cloud
- 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.)
- Active
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
-
- 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)
- Length Measuring Devices By Optical Means (AREA)
- Image Processing (AREA)
Abstract
The present invention relates to automobile assistant driving or automatic Pilot technical fields, efficient road surface and curb detection method more particularly to a kind of Lidar sensor based on information fusion, the present invention filters filter method using unique two-wire, the local detail information (and these details are for accurately identifying that curb and road surface have very important effect) that a cloud is reflected is remained while filtering out sharp noise and outlier, the extraction for subsequent high-precision road surface curb provides solid foundation.The present invention can not only realize stable and excellent road surface curb testing result in good urban road environment, and can equally detect more intact road surface and curb under much noise, barrier and complicated road conditions.Meanwhile this method not only has very strong detectability to the Lidar point cloud of high harness, but also to low line number Lidar point cloud information, equally has good detectability.
Description
Technical field
The present invention relates to automobile assistant driving or automatic Pilot technical fields, and in particular to a kind of L based on information fusion
The efficient road surface of idar sensor and curb detection method.
Background technique
In automatic Pilot field, curb and pavement detection are to form one of the element task in travelable region, to realization L4
The automatic Pilot of rank has vital effect.In traditional solution based on 2D vision, due to having lacked depth
Information is spent, is affected under the scene of (such as Qiang Guang, night) in certain visual informations, road surface and curb detection can not obtain well
To solution.In the 3D sensor using Lidar as main sensors, the available good solution of this problem.However, by
In by sensor at this stage development and scientific research level limited, extracted from 3D point cloud healthy and strong and stable road surface and
Curb information or a difficult point.
In general, the point cloud density of the Lidar used in automatic Pilot scene is relatively low, and the feature of curb exists
In local detail, meanwhile, these local details are easy to be influenced by a large amount of sharp noises and outlier again.Therefore, how
The minutia that curb is obtained from low-density point cloud information is a difficult point of current road curb detection.And it uses existing
Filter (common filter such as arithmetic mean filter, median filter, Gaussian filter) filters sharp noise and peels off
When point, the loss of local detail often will cause, and these details are exactly to extract the key of curb information.
Summary of the invention
Mesh of the invention it is to solve problem of the prior art, a kind of Lidar sensor based on information fusion is provided
Efficient road surface and curb detection method, healthy and strong, stable road surface and curb characteristic information are got, to realize road pavement
With the high efficiency extraction of curb information.
To achieve the above object, design it is a kind of based on information fusion Lidar sensor efficient road surface and curb detection
Method, including a cloud separated time processing method, filtering filtration treatment method, carry out a cloud characteristic differentiation and integral operation method,
The method for carrying out road surface, curb and barrier fusion method and calculating road surface principal direction, it is characterised in that at the filtering filtering
Reason filters filter method using two-wire, and the specific method is as follows:
First step carries out line internal sort processing for every point cloud line, this is ordered as according to it in original point cloud
Index the sequence of size progress from big to small;
Second step is to every point cloud line, since the smallest index point, to each point forward detection its in this line remain
A nearest point of Euclidean distance in the set that remaining untreated point is constituted, the starting point by retaining and as detection next time,
Simultaneously current point is identified as it is processed, until being disposed, this processing retain point be first line filter result;
Third step is to every point cloud line, since the point of largest index, to each point after to detect its in this line remain
A nearest point of Euclidean distance in the set that remaining untreated point is constituted, the starting point by retaining and as detection next time,
Current point is identified as processed simultaneously, until being disposed, this is the filter result of Article 2 line;
4th step merges the point cloud that second step and third step acquire, and traverses and obtains in second step and third step
Point, the point all retained in this two step is then considered as feature stable point, needs to remain, these points finally obtained are
The filter result of two-wire filter.
A kind of efficient road surface of Lidar sensor based on information fusion and curb detection method, feature exist
It is specific as follows in described cloud separated time processing method: according to lidar sensor coordinate system position and with the scan line institute on ground
The geometrical relationship of formation carries out separated time to the point cloud got, which includes: every scan line and the ground of sensor
Angle, and scanned point cloud in each point the successive positional relationship of index.
A kind of efficient road surface of Lidar sensor based on information fusion and curb detection method, feature exist
It is specific as follows in the characteristic differentiation and integral operation method for carrying out a cloud: to third step treated point cloud in each point carry out office
Portion's elevation calculus of differences carries out integral operation to resulting elevation difference information, and in the result of acquisition, according to road later
A cloud is filtered along the elevation of, road surface and barrier, scale size and point cloud density feature, realize curb, road surface and
Divide between cluster and class in the preliminary class of barrier.
A kind of efficient road surface of Lidar sensor based on information fusion and curb detection method, feature exist
In carrying out, road surface, curb and barrier fusion method are specific as follows: line segment is gone out to local point-cloud fitting by least square method, and
The angle and curvature for estimating partial points cloud respectively according to line segment angle and variable angle rate, in combination with what is obtained in third step
Integral elevation information and point Yun Midu, to third step obtain road surface, curb and barrier merge.
A kind of efficient road surface of Lidar sensor based on information fusion and curb detection method, feature exist
It is specific as follows in the method for calculating road surface principal direction: using PCA method road pavement, curb and barrier in machine learning into
Row filtering, filter process are as follows: with the consistent road surface of principal direction be considered as effective road surface, the curb on effective road surface is considered as having
Curb is imitated, by filtering, finally extracts high-precision and the road surface semantically standardized and curb.
The present invention filters filter method using unique two-wire, and a cloud is remained while filtering out sharp noise and outlier
The local detail information (and these details are for accurately identifying that curb and road surface have very important effect) reflected is
The extraction of subsequent high-precision road surface curb provides solid foundation.Stabilization can not only be realized in good urban road environment
And excellent road surface curb testing result, and can equally be detected under much noise, barrier and complicated road conditions
More intact road surface and curb.Meanwhile this method not only has very strong detectability to the Lidar point cloud of high harness, but also
To low line number Lidar point cloud information, equally there is good detectability.This method is possessed in NULLMAX
95% recall rate and 92% accuracy rate are achieved on data set.
Detailed description of the invention]
Fig. 1 is unfiltered original image；
Fig. 2 is the partial schematic diagram of Fig. 1；
Fig. 3 is first time processing result figure；
Fig. 4 is second of processing result figure；
Fig. 5 is Fig. 3 and Fig. 4 amalgamation result figure；
Fig. 6 is final result figure；
Fig. 7 is that point cloud separated time handles geometrical relationship figure.
In figure: 1, starting point.
Specific embodiment
The technical scheme of the present invention will be further described with reference to the accompanying drawings and embodiments, and following instance is for illustrating this
Invention, but cannot be used to limit the scope of the invention.
Embodiment:
One, point cloud separated time processing: referring to Fig. 7
Point cloud separated time processing is specific as follows:
1, the horizontal sextant angle between the point and Lidar sensor in same scanning loop wireIt is identical；
2, the horizontal sextant angle between the point and Lidar sensor of different scanning loop wireIt is different；
To the point cloud point got, its horizontal folder between Lidar sensor is calculated according to the above-mentioned geometrical relationship fact
Angle, because the point of same loop wire is identical as the horizontal sextant angle between Lidar sensor, and the point and Lidar sensor of different loop wires
Horizontal sextant angle it is different.It is configured as a result, according to the angle between Lidar equipment difference loop wire, the point cloud point that will acquire is by loop wire
Carry out separated time.It is important to note that the processing of following third, four steps is carried out inside each loop wire, and the 5th
Step is carried out on treated whole point cloud.
Two, two-wires filter filter method:
For vivider explanation, the present embodiment provides that represent the treatment process in region in a wherein ring of the invention to illustrate
Validity:
Referring to Fig. 1, Fig. 1 is that the point in unfiltered original image and Fig. 1 is ranked good, it is clockwise by starting point 1
The index value of the point in direction since 0 and is incremented by.
1, the click-through line internal sort in every point cloud line is handled.This is ordered as according to it in original point cloud
Indexing the sequence that size carries out from small to large is referring to fig. 2 one section in Fig. 1.
2, to every point cloud line, if the point that index is i is pi.Since indexing the point for 0, forward detection current point to one
The closest approach is retained and is detected as next time by the nearest point of Euclidean distance in the set that untreated point is constituted in a window k
Starting point, while current point is identified as it is processed, until being disposed.The point that this processing retains is the filtering of first line
As a result.For example, k=4 are set, current point pi, then p is calculatediTo p(i+1), p(i+2), p(i+3), p(i+4)Middle Euclidean distance is nearest
Point.If p(i+1)For distance piClosest approach, then by piRetain and be labeled as processed, and current point is set as p(i+3), calculating should
Point arrives p(i+4), p(I+5), p(i+6), p(i+7)The nearest point of middle Euclidean distance, so circulation is gone down until this line is disposed.This
Effect signal is walked referring to Fig. 3, is the result figure of step 2, overstriking point therein represents this step result.
3, to every point cloud line, since the point of largest index, to after each point to detection current point to a window k
A nearest point of Euclidean distance in the set that interior untreated point is constituted, the starting point by retaining and as detection next time,
Simultaneously current point is identified as it is processed, until being disposed.This is the filter result of Article 2 line.For example, setting k=4, currently
Point is pi, then p is calculatediTo p(i-1), p(i-2), p(i-3), p(i-4)The nearest point of middle Euclidean distance.If p(i-3)For distance piIt is nearest
Point, then by piRetain and be identified as processed, and current point is set as p(i-3), the point is calculated to p(i-4), p(i-5), p(i-6), p(i-7)
The nearest point of middle Euclidean distance, so circulation is gone down until this line is disposed.It referring to fig. 4, is the result figure of step 3,
Wherein thick point represents this step result.
4, to every point cloud line, the point cloud that step 2 and step 3 retain is merged.It is protected in traversal step 2 and step 3
The point stayed, the point all retained in this two step are then considered as feature stable point, need to retain.These points finally obtained are double
The filter result of line filter.Referring to Fig. 5, to merge the filter result that Fig. 3, Fig. 4 common ground are formed.Referring to Fig. 6, for most
Whole result figure.
Three, carry out characteristic differentiation and the integral operation of a cloud:
To second step treated point cloud, following operation is executed:
1. calculating each cloud point pmDepth displacement in the neighborhood that size is n, formula are as follows:
Wherein m is the index of current point, dmFor depth displacement, pm+iIt .z is the z value for the point that index is m+i, pm+i-1It .z is index
For the z value of the point of m+i-1, z is the elevation relative to datum level.
2. being clustered after pair process step 1 processing according to the depth displacement d of point each in cloud.Wherein, meet poly-
The point cloud of class condition d < ρ (ρ is elevation differential threshold) is Local Road Surface candidate point cloud, calculates the mean height in this partial points cloud
Journey, being greater than h(h if dispersed elevation is elevation threshold value) if this Local Road Surface candidate point cloud given up；And meet cluster condition d >=ρ
Point cloud be local curb candidate point cloud, calculate the accumulation depth displacement D in this local curb candidate point cloudhWith line density De/ n(its
Middle DeRefer to the sum of the adjacent two o'clock Euclidean distance of part curb candidate point cloud:, wherein d (pi,
p(i+1)) it is to calculate piAnd p(i+1)Between Euclidean distance function, n is the quantity of local curb candidate point cloud), if this local road
Depth displacement D is accumulated along candidate point cloudhIt is unsatisfactory for being greater than plAnd it is less than pu(in general, the setting of p is less than 10cm, puSetting
Greater than 30cm) or line density De/ n is less than threshold value, then by this part, curb candidate point cloud is given up.
The fusion on four, progress road surface, curb:
1. merging adjacent Local Road Surface candidate point cloud in every point cloud line.To two adjacent parts in cloud line
Road surface candidate point cloud C1And C2(the adjacent indexed sequential referred to according to point herein, C1And C2Between there is no other Local Road Surface candidate
Point cloud), to C1With C2W point in mutual nearest one end of Euclidean distance is (assuming that C1In point index be less than C2In point
Index, then point herein refers to C1The maximum w point of middle index value and C2The middle the smallest w point of index value), use least square
Optimization fits line segment l respectively1And l2, and according to line segment l1And l2Angle and depth displacement waited to estimate adjacent Local Road Surface
The angle and depth displacement of reconnaissance cloud adjoiner, if l1And l2Angle be less than threshold value and depth displacement and be less than threshold value, then it is this is adjacent
The corresponding Local Road Surface candidate point cloud fusion of line segment, becomes new Local Road Surface candidate point cloud, and calculates updated local road
The dispersed elevation of face candidate point cloud.This Local Road Surface candidate point cloud is given up if dispersed elevation is greater than threshold value.
2. merging adjacent local curb candidate point cloud in every point cloud line.Using least square method to adjacent part
Curb candidate point cloud C1And C2(the adjacent indexed sequential referred to according to point herein, C1And C2Between there is no other local curb candidate
Point cloud or Local Road Surface candidate point cloud) difference matching line segment l1And l2.According to line segment l1And l2Angle and distance judged,
If l1And l2Angle be less than threshold value and distance is less than threshold value, then this adjacent local curb candidate point cloud is merged, becomes new
Local curb candidate point cloud.Judged according to the accumulation depth displacement of this updated local curb candidate point cloud: if
Accumulate depth displacement DhIt is unsatisfactory for being greater than plLess than puOr line density De/ n is less than threshold value, then by this part, curb candidate point cloud is given up.
3. further being judged according to adjacent road surface and curb.To adjacent Local Road Surface candidate point cloud CrDrawn game
Portion curb candidate point cloud Cc(the adjacent indexed sequential referred to according to point herein, CrAnd CcBetween there is no other local curb candidate point
Cloud or Local Road Surface candidate point cloud) the nearest one end of mutual Euclidean distance in n point (assuming that CrIn the index of point be less than
CcIn point index, then point herein refers to CrThe maximum n point of middle index value and CcThe middle the smallest n point of index value) respectively
Calculate its elevation hrAnd hcIf hrGreater than hcThen give up this Local Road Surface candidate point cloud.
This step further improves the anti-interference ability to noise, while the road surface extracted, and curb semantic feature is more
Add obvious.
Five, extract road surface principal direction, and filter out main road face
The whole Local Road Surface candidate point clouds that will be obtained by the 4th step, are mapped on X/Y plane and (map and refer to herein
Neglect the value of Z in a cloud), it is calculated using a kind of principal component analytical method (PCA, common statistical analysis algorithm) scheduling algorithm
Its principal direction is obtained, this principal direction is the direction V on whole road surfacep.Constructed origin o(0,0) and with VpFor the straight line in direction
lp, each Local Road Surface candidate point cloud is made the following judgment on X/Y plane: if each Local Road Surface candidate point cloud and lpPhase
Friendship or and lpBetween angle be less than threshold value, then the Local Road Surface candidate point cloud become Local Road Surface point cloud.Wherein, Local Road Surface
Candidate point cloud and lpIntersection refers to two endpoint e of Local Road Surface candidate point cloudlAnd erThe triangle and straight line formed with origin o
lpIntersection point fall in elAnd erOn the line segment of composition；And Local Road Surface candidate point cloud and lpBetween angle refer to straight line oelAnd oer
Respectively with lpThe minimum value in two angles formed.
The local curb candidate point cloud obtained in the 4th step is filtered according to the Local Road Surface of acquisition point cloud, with part
Road surface point cloud CrAdjacent local curb candidate point cloud CcThen become local curb point cloud (the adjacent index referred to according to point herein
Sequentially, CrAnd CcBetween there is no other local curb candidate point cloud or Local Road Surface candidate point cloud).By the filtering of this step, finally
Extract the high-precision road surface semantically standardized and curb.This filter process can summarize are as follows: with the consistent road surface of principal direction
It is considered as effective road surface, the curb for limiting effective road surface is considered as effective curb.
Claims (4)
1. efficient road surface and the curb detection method of a kind of Lidar sensor based on information fusion, including cloud separated time processing
Method, filtering filtration treatment method, the characteristic differentiation for carrying out a cloud and integral operation method carry out road surface, curb and barrier
Fusion method and the method for calculating road surface principal direction, it is characterised in that the filtering filtration treatment filters filtering side using two-wire
Method, method are specific as follows:
First step carries out line internal sort processing for every point cloud line, this is ordered as according to its index in original point cloud
Size carries out sequence from big to small；
Second step is to every point cloud line, since the smallest index point, to each point forward detection its in this line it is remaining not
The point that Euclidean distance is nearest in the set constituted is put in processing, by reservation and as the next starting point detected, simultaneously
Current point is identified as it is processed, until being disposed, this processing retain point be first line filter result；
Third step is to every point cloud line, since the point of largest index, to after each point to detect its in this line it is remaining not
The point that Euclidean distance is nearest in the set constituted is put in processing, by reservation and as the next starting point detected, simultaneously
Current point is identified as processed, until being disposed, this is the filter result of Article 2 line；
4th step merges the point cloud that second step and third step acquire, and traverses point obtained in second step and third step,
The point all retained in this two step is then considered as feature stable point, needs to remain, these points finally obtained are two-wires
The filter result of filter；
Characteristic differentiation and the integral operation method for carrying out a cloud are specific as follows: carrying out to each point in third step treated point cloud
Local elevation calculus of differences carries out integral operation to resulting elevation difference information, and in the result of acquisition later, according to
Curb, the elevation on road surface and barrier, scale size and point cloud density feature a cloud is filtered, realize curb, road surface with
And divide between cluster and class in the preliminary class of barrier.
2. the efficient road surface and curb detection side of a kind of Lidar sensor based on information fusion as described in claim 1
Method, it is characterised in that described cloud separated time processing method is specific as follows: according to lidar sensor coordinate system position and and ground
Scan line be formed by geometrical relationship, separated time is carried out to the point cloud that gets, which includes: that every of sensor sweeps
The angle on line and ground is retouched, and the successive positional relationship of index for putting each point in cloud scanned.
3. the efficient road surface and curb detection side of a kind of Lidar sensor based on information fusion as described in claim 1
Method, it is characterised in that it is specific as follows to carry out road surface, curb and barrier fusion method: quasi- to partial points cloud by least square method
Line segment is closed out, and estimates the angle and curvature of partial points cloud respectively according to line segment angle and variable angle rate, in combination with point
The integral elevation information and point Yun Midu obtained in the characteristic differentiation and integral operation method of cloud, road surface, curb to acquisition
And barrier merges.
4. the efficient road surface and curb detection side of a kind of Lidar sensor based on information fusion as described in claim 1
Method, it is characterised in that the method for calculating road surface principal direction is specific as follows: PCA method road pavement, curb in machine learning are utilized
And barrier is filtered, filter process are as follows: with the consistent road surface of principal direction is considered as effective road surface, the curb on effective road surface
It is considered as effective curb, by filtering, finally extracts high-precision and the road surface semantically standardized and curb；
It is specific as follows: the whole Local Road Surface candidate point clouds obtained by the 4th step being mapped on X/Y plane, the side PCA is utilized
Method, which calculates, obtains its principal direction, and principal direction is the direction Vp on whole road surface, constructs origin o(0, and 0) and using Vp as direction
Straight line lp makes the following judgment each Local Road Surface candidate point cloud on X/Y plane: if each Local Road Surface candidate point cloud with
Lp intersection or the angle between lp are less than threshold value, then the Local Road Surface candidate point cloud becomes Local Road Surface point cloud；Wherein, part
Road surface candidate point cloud intersect with lp refer to Local Road Surface candidate point cloud two endpoint el and er and origin o formed triangle with
The intersection point of straight line lp is fallen on the line segment that el and er is constituted；And the angle between Local Road Surface candidate point cloud and lp refers to straight line
The minimum value in two angles that oel and oer is formed with lp respectively.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910169999.5A CN109614966B (en) | 2019-03-07 | 2019-03-07 | It is a kind of based on information fusion Lidar sensor efficient road surface and curb detection method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910169999.5A CN109614966B (en) | 2019-03-07 | 2019-03-07 | It is a kind of based on information fusion Lidar sensor efficient road surface and curb detection method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109614966A CN109614966A (en) | 2019-04-12 |
| CN109614966B true CN109614966B (en) | 2019-07-26 |
Family
ID=66019135
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910169999.5A Active CN109614966B (en) | 2019-03-07 | 2019-03-07 | It is a kind of based on information fusion Lidar sensor efficient road surface and curb detection method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109614966B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110163880B (en) * | 2019-06-04 | 2020-04-14 | 奥特酷智能科技(南京)有限公司 | Method for acquiring point cloud road surface height in Unity |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108681525A (en) * | 2018-05-16 | 2018-10-19 | 福州大学 | A kind of road surface point cloud intensity enhancing method based on Vehicle-borne Laser Scanning data |
| CN109101743A (en) * | 2018-08-28 | 2018-12-28 | 武汉市众向科技有限公司 | A kind of construction method of high-precision road net model |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| GB2532948B (en) * | 2014-12-02 | 2021-04-14 | Vivo Mobile Communication Co Ltd | Object Recognition in a 3D scene |
-
2019
- 2019-03-07 CN CN201910169999.5A patent/CN109614966B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108681525A (en) * | 2018-05-16 | 2018-10-19 | 福州大学 | A kind of road surface point cloud intensity enhancing method based on Vehicle-borne Laser Scanning data |
| CN109101743A (en) * | 2018-08-28 | 2018-12-28 | 武汉市众向科技有限公司 | A kind of construction method of high-precision road net model |
Non-Patent Citations (4)
| Title |
|---|
| Curvature-based curb detection method in urban environments using stereo and laser;C. Fernández 等;《2015 IEEE Intelligent Vehicles Symposium》;20150827;第579-584页 |
| Road-Segmentation-Based Curb Detection Method for Self-Driving via a 3D-LiDAR Sensor;Yihuan Zhang 等;《IEEE Transactions on Intelligent Transportation Systems》;20181231;第19卷(第12期);第3981-3991页 |
| 基于曲率统计的LiDAR点云二次滤波方法;万剑华 等;《中国石油大学学报(自然科学版)》;20130228;第37卷(第1期);第56-60页 |
| 机载三维激光扫描点地面点剔除算法;第4期;《大地测量与地球动力学》;20090830;第29卷(第4期);第97-101页 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109614966A (en) | 2019-04-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109684921B (en) | Road boundary detection and tracking method based on three-dimensional laser radar | |
| CN105866790B (en) | A kind of laser radar obstacle recognition method and system considering lasing intensity | |
| Zhou et al. | Seamless fusion of LiDAR and aerial imagery for building extraction | |
| CN105844629B (en) | A kind of large scene City Building facade point cloud automatic division method | |
| Heipke et al. | Evaluation of automatic road extraction | |
| CN108197583B (en) | Building change detection method based on graph cut optimization and image structure characteristics | |
| CN106780524A (en) | A kind of three-dimensional point cloud road boundary extraction method | |
| CN112184736B (en) | Multi-plane extraction method based on European clustering | |
| CN105973265B (en) | A kind of mileage estimation method based on scanning laser sensor | |
| CN108898672A (en) | A kind of semi-automatic cloud method making three-dimensional high-definition mileage chart lane line | |
| CN111325138B (en) | Road boundary real-time detection method based on point cloud local concave-convex characteristics | |
| CN106595659A (en) | Map merging method of unmanned aerial vehicle visual SLAM under city complex environment | |
| CN101604450A (en) | The method of integrated images and LiDAR data extract contour of building | |
| CN111178150B (en) | Lane line detection method, system and storage medium | |
| JP2008065634A (en) | Object detection apparatus and object detection method | |
| CN106355197A (en) | Navigation image matching filtering method based on K-means clustering algorithm | |
| US20220398856A1 (en) | Method for reconstruction of a feature in an environmental scene of a road | |
| WO2022121177A1 (en) | Scan line-based road point cloud extraction method | |
| Awrangjeb et al. | Rule-based segmentation of LIDAR point cloud for automatic extraction of building roof planes | |
| CN112166457A (en) | Point cloud segmentation method and system and movable platform | |
| CN114782729A (en) | Real-time target detection method based on laser radar and vision fusion | |
| CN113538501A (en) | Low-altitude image DSM generation building edge refinement method | |
| CN109614966B (en) | It is a kind of based on information fusion Lidar sensor efficient road surface and curb detection method | |
| Rothermel et al. | Fast and robust generation of semantic urban terrain models from UAV video streams | |
| CN111507341A (en) | Method, device and equipment for adjusting target bounding box and storage medium |
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 |