CN108759667B - Front truck distance measuring method under vehicle-mounted camera based on monocular vision and image segmentation - Google Patents

Front truck distance measuring method under vehicle-mounted camera based on monocular vision and image segmentation Download PDF

Info

Publication number
CN108759667B
CN108759667B CN201810529462.0A CN201810529462A CN108759667B CN 108759667 B CN108759667 B CN 108759667B CN 201810529462 A CN201810529462 A CN 201810529462A CN 108759667 B CN108759667 B CN 108759667B
Authority
CN
China
Prior art keywords
vehicle
image
information
bounding box
front truck
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.)
Expired - Fee Related
Application number
CN201810529462.0A
Other languages
Chinese (zh)
Other versions
CN108759667A (en
Inventor
黄立勤
陈雅楠
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Fuzhou University
Original Assignee
Fuzhou University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Fuzhou University filed Critical Fuzhou University
Priority to CN201810529462.0A priority Critical patent/CN108759667B/en
Publication of CN108759667A publication Critical patent/CN108759667A/en
Application granted granted Critical
Publication of CN108759667B publication Critical patent/CN108759667B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
    • G01B11/00—Measuring arrangements characterised by the use of optical techniques
    • G01B11/02—Measuring arrangements characterised by the use of optical techniques for measuring length, width or thickness
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
    • G01B11/00—Measuring arrangements characterised by the use of optical techniques
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
    • G01B11/00—Measuring arrangements characterised by the use of optical techniques
    • G01B11/24—Measuring arrangements characterised by the use of optical techniques for measuring contours or curvatures
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/50—Depth or shape recovery
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/90—Determination of colour characteristics

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Geometry (AREA)
  • Computer Graphics (AREA)
  • Software Systems (AREA)
  • Image Analysis (AREA)
  • Traffic Control Systems (AREA)

Abstract

The present invention discloses the front truck distance measuring method based on monocular vision and image segmentation under a kind of vehicle-mounted camera, is primarily based on the extraction that deep learning algorithm carries out two-dimentional bounding box and three-dimensional boundaries frame to target vehicle, obtains its corresponding position information；Secondly, the length, width and height based on three-dimensional boundaries frame are matched with 3D CAD auto model, the corresponding approximate three-dimensional vehicle model of vehicle is obtained；Again and, two-dimentional bounding box is based on to vehicle extraction vehicle classification information in figure；Therewith, the corresponding three-dimensional information of vehicle and vehicle model information are sent into example segmentation network, according to camera imaging principle, the absolute depth values of vehicle in image is calculated according to the dimension information of different automobile types.It present invention saves the time of depth calculation and ensure that driving vision, driver allow intuitively to observe the specific distance value of front truck, make reasonable drive and judge.

Description

Front truck distance measuring method under vehicle-mounted camera based on monocular vision and image segmentation
Technical field
The present invention relates to technical field of computer vision, and in particular to based on monocular vision and figure under a kind of vehicle-mounted camera As the front truck distance measuring method of segmentation.
Background technique
Automobile universal is so that demand of the world to intelligent vehicle increasingly increases, and the development of machine vision, allows machine Image information can be generally obtained with human eye.
Estimation of Depth based on video image is to realize the precondition of Vehicular automatic driving and safe driving, so-called image Estimation of Depth be the two dimensional image stream obtained according to video equipment acquisition of information real world under correspond to the actual range of object Method.Traditional distance measuring method needs to demarcate camera inside and outside parameter with camera heights mostly, currently based on nerve net The distance measuring method of network study integrally carries out relative depth measurement mainly for image.
Common distance measuring method neural network based mainly designs different convolutional neural networks structures at this stage, instructs The loss function for practicing known depth image obtains depth model, directly obtains to the image of input in test corresponding relatively deep Degree figure.Such method indicates relative depth with color, it is necessary to for obtaining the demand of depth globality, as indoor ranging this It is that there are some superiorities, but for actual traffic scene, redundancy and traffic are travelled from the point of view of the closed spatial dimension of kind Correlation is weaker in the process, it is little to detect its depth information advantage in open space, or even can reduce detection efficiency, and Indicating far and near with color has that comparison color change is small, is difficult to judge that vehicle is specifically far and near.When measured object with take the photograph Depth can be indicated significantly by color when camera is closer, but for object remotely, due to color change width Degree is big, be visually difficult to recognize the specific depth information of distant objects and when being overlapped there are more vehicle or vehicle not Can completely provides the information of vehicle.It, can not using the differential expression relative depth information of color under true traffic scene Real safety guarantee is provided to user.
Under true traffic scene, traditional distance measuring method estimates picture depth merely with geometric projection relationship Meter, such methods, which need to demarcate distinct device, is readily incorporated error, and the side for carrying out ranging using laser radar instrument Method, since distance detectable under specific place is limited and needs to introduce new equipment in cost and convenience It sees, such method is not the best approach of current ranging.And the common distance measuring method of neural network is used to image entirety ranging And indicate that the distant relationships of vehicle in image exist using color difference, not only intuitive range information cannot be provided to driver Its emergency reaction time to emergency is increased, while being also unfavorable for the observation to the road ahead visual field.
Making a general survey of the existing depth estimation method for being mostly based on neural network all has two: 1) estimating relative depth Rather than absolute depth；It 2) include bulk redundancy item, such as sky, the building of distant place and two sides shade tree etc. in depth information.
Summary of the invention
In view of the deficiencies of the prior art, the present invention is provided under a kind of vehicle-mounted camera based on monocular vision and image segmentation Front truck distance measuring method solves the vehicle real time distance being suitable under traffic scene, only detects the vehicle under traffic scene, does not utilize Color difference indicates relative distance, and intuitively provides absolute distance of each car in true.
To achieve the above object, the technical scheme is that being based on monocular vision and image under a kind of vehicle-mounted camera The front truck distance measuring method of segmentation, comprising the following steps:
Step S1: image is read by frame from the video flowing that vehicle-mounted camera is shot；
Step S2: target detection is carried out to vehicle, extracts the location information of vehicle in image, including two-dimentional bounding box information With three-dimensional boundaries frame information；
Step S3: the CAD of corresponding vehicle is matched in preset vehicle 3D CAD model library according to three-dimensional boundaries frame information Model；
Step S4: the distance of overlapping vehicle is judged according to the two-dimentional bounding box information of vehicle, according to vehicle two dimension bounding box Top left corner apex ordinate value, CAD model projection exposure mask on the image determines the block information of vehicle；Step S5: according to The two-dimentional bounding box information of vehicle detects vehicle, obtains the dimension information of vehicle according to specific vehicle；
Step S6: corresponding position in image is obtained using example segmentation network according to the dimension information of vehicle and block information The Pixel-level exposure mask value of vehicle；
Step S7: according to camera imaging principle, each car is calculated using the dimension information of vehicle and the result of vehicles segmentation Corresponding absolute depth values and the Some vehicles absolute depth values that are blocked.
Further, the step S3 is specifically included:
According to three-dimensional boundaries frame information, chosen in preset vehicle 3D CAD model library and the cuboid of target vehicle 3D CAD model mapping is matched in the 3D cuboid of respective objects vehicle by the immediate model of length and width high proportion.
Further, the step S4 is specifically included: in image coordinate system, comparing the upper left of the two-dimentional bounding box of vehicle Angular vertex ordinate value, the small corresponding vehicle of two-dimentional bounding box of ordinate value are nearby vehicle, the big two-dimentional side of ordinate value Frame corresponding vehicle in boundary's is remote vehicle, and according to the top left corner apex ordinate value of vehicle two dimension bounding box, CAD model is projected in Exposure mask on image determines the visible part of target vehicle, the part being blocked and the part being truncated by image.
Further, the vehicle includes multi-functional Recreational Vehicle, sport vehicle, wing-rooms on either side of a one-story house type hatchback vehicle, four Car, mini bus, streamlined car, four three compartment extended type station wagons, picking-up vehicle, transboundary Recreational Vehicle, open car, sport car and steel Plate top sport car.
Further, the step S7 is specifically included:
According to camera imaging principle, fathomed using the relationship between the elemental area of vehicle and true area, In In image coordinate system, the coordinate of the two-dimentional bounding box top left corner apex of vehicle is q (x, y, f), in actual spatial coordinates system, vehicle Two-dimentional bounding box top left corner apex coordinate be Q (X, Y, Z), then
It is write as matrix form, is obtained
Pixel coordinate system is defined as (u, v), Q1(u0,v0) be image coordinate system plane and optical axis intersection point, image coordinate Point coordinate in system is (x, y), is reflected as in pixel coordinate system (u, v), then image coordinate system and pixel coordinate system exist with Lower relationship, wherein dxdy indicates the physics size of each pixel
It is write above formula (3) as matrix form, is obtained
Formula (2) is brought into formula (4), is obtained
Do not consider translation problem, enables u0=v0=0,fx=fy=7.1254 × 102
Dividing vehicle area is S under pixel coordinate systempixel
It is S that dividing vehicle area is wherein corresponded under real spacereal=xy, Z axis coordinate value correspond to take the photograph under real space Distance d as head apart from front truck, obtains
Compared with prior art, the present invention have the utility model has the advantages that
(1) present invention combines example to divide to obtain corresponding vehicle location and shape in image and believes using objective detection Breath, eliminates the interference of redundancy in detection, this scheme saves the time of depth calculation and ensure that driving vision, gives Driver provides longer operable time.
(2) monocular vision principle is utilized, according to objective detection vehicle under actual time in conjunction with example dividing method Dimensioned area directly calculates the absolute depth values of corresponding vehicle and provides specific value information, do not utilize color is fuzzy to distinguish Distant relationships, this scheme allow driver intuitively to observe the specific distance value of front truck, make reasonable drive and judge.
Detailed description of the invention
Fig. 1 is the process of the front truck distance measuring method based on monocular vision and image segmentation under a kind of vehicle-mounted camera of the present invention Schematic diagram；
Fig. 2 is the schematic diagram of judgement overlapping headlight for vehicle of the invention；
Fig. 3 is the schematic diagram of occlusion information of the present invention；
Fig. 4 is that example divides network Mask R-CNN structural schematic diagram in one embodiment of the invention；
Fig. 5 is vehicle classification schematic diagram in one embodiment of the invention；
Fig. 6 is projection relation schematic diagram in one embodiment of the invention；
Fig. 7 is the plane of delineation and pixel planes transition diagram in one embodiment of the invention.
Specific embodiment
The present invention will be further described with reference to the accompanying drawings and embodiments.
As shown in Figure 1, the front truck distance measuring method under a kind of vehicle-mounted camera based on monocular vision and image segmentation, including with Lower step:
Step S1: image is read by frame from the video flowing that vehicle-mounted camera is shot；
Step S2: target detection is carried out to vehicle, extracts the location information of vehicle in image, including two-dimentional bounding box information With three-dimensional boundaries frame information；
Step S3: the CAD of corresponding vehicle is matched in preset vehicle 3D CAD model library according to three-dimensional boundaries frame information Model；
Step S4: the distance of overlapping vehicle is judged according to the two-dimentional bounding box information of vehicle, according to vehicle two dimension bounding box Top left corner apex ordinate value, CAD model projection exposure mask on the image determines the block information of vehicle；
Step S5: detecting vehicle according to the two-dimentional bounding box information of vehicle, is believed according to the size that specific vehicle obtains vehicle Breath；
Step S6: corresponding position in image is obtained using example segmentation network according to the dimension information of vehicle and block information The Pixel-level exposure mask value of vehicle；
Step S7: according to camera imaging principle, each car is calculated using the dimension information of vehicle and the result of vehicles segmentation Corresponding absolute depth values and the Some vehicles absolute depth values that are blocked.
Using monocular vision basis, in conjunction with objective detection with image instance divide measurement road in front vehicles it is exhausted To depth.
In the present embodiment, video flowing is extracted in image according in frame sequential input network by target detection network The location information of vehicle, including two-dimentional bounding box and three-dimensional boundaries frame information as the subsequent prerequisite item for calculating front truck absolute depth Part.For objective detection part, the corresponding 3D CAD model of each vehicle is stored in the model library of web in image The model set that (Trimble 3D Warehouse) is pre-defined, according to the three-dimensional boundaries detected from 2D image Frame location information most connects to choose in the data acquisition system of 3D CAD model with the length and width of the cuboid of target vehicle at high proportion Most like 3D CAD model is reflected and is matched in the 3D cuboid of respective objects vehicle by close model.It is remote to overlapping vehicle Nearly judgement, as shown in Fig. 2, comparing two top left co-ordinate y values of two-dimentional bounding box, the corresponding vehicle of the small bounding box of y value is Be blocked vehicle position nearby vehicle, then the big bounding box of y value corresponds to remote vehicle.According to the upper left corner of vehicle two dimension bounding box Y value size, CAD model projection exposure mask on the image can determine the visible part of target vehicle, be blocked and be truncated part As shown in Figure 3.
In the present embodiment, utilize Mask R-CNN network to figure two dimension target detection part and example partitioning portion As carrying out processing operation.Mask R-CNN is one and divides network based on the example that Faster R-CNN extends, for each mesh Mark object has three output branchs, is classification branch, bounding box branch and mask branch respectively, as shown in Figure 4.Utilize Mask R-CNN first detects target position (i.e. the output of bounding box branch), and the principle of cutting object, is not repeatedly introduced target detection later Network and be completed at the same time the extraction of target position and the segmentation of target using Mask R-CNN network.Example partitioning portion utilizes The parted pattern only for vehicle of re -training carries out the acquisition of result, it is ensured that does not occur in detection process except other than vehicle Redundancy.
Rear end is conveyed to by the vehicle position information that Mask R-CNN network obtains and carries out example cutting operation, is conveyed simultaneously Vehicle classification network is given, the vehicle for corresponding to vehicle location in image is extracted.Vehicle detection part uses Compcars data Collect trained vehicle classification model.Vehicle is divided into 12 classes, as shown in figure 5, respectively multi-functional Recreational Vehicle (Multi- Purpose Vehicle,MPV)；Sport vehicle (Sport Utility Vehicle, SUV)；Wing-rooms on either side of a one-story house type hatchback vehicle (hatchback)；Four-door sedan (sedan)；Mini bus (minibus)；Streamlined car (fastback)；Four three compartments lengthen Type station wagon (estate)；Picking-up vehicle (pickup)；Recreational Vehicle (crossover) transboundary；Open car (convertible)；It runs Vehicle (sport)；Steel plate top sport car (hardtop convertible).Vehicle in figure is obtained according to resulting vehicle position information It is sequentially sent to obtain corresponding vehicle model information in vehicle classification network, obtains the corresponding length, width and height of vehicle further according to specific vehicle and believe Breath, so as to improve calculated absolute depth accuracy, model data is as shown in table 1.
Table 1
On the basis of the detection of the above objective, example segmentation are with vehicle classification, solved by distance measuring method below Certainly it is present in the problem of depth definition in estimation of Depth is with detection redundancy.In the absolute depth for obtaining each vehicle while reducing The appearance of redundancy.According to camera imaging principle, location algorithm can be reduced to such as drag, such as Fig. 6 and Fig. 7.
It according to camera imaging principle, is fathomed using the relationship between elemental area and true area, (x, y) is figure As the coordinate of coordinate system, (X, Y, Z) is the coordinate of actual spatial coordinates system, as shown in Figure 6.
It is write above formula (1) as matrix form, is obtained
Such as Fig. 7, pixel coordinate system is defined as (u, v), Q1(u0,v0) be image coordinate system plane and optical axis intersection point.Figure As the point coordinate in coordinate system be (x, y), be reflected as in pixel coordinate system (u, v).Then image coordinate system and pixel coordinate system There are following relationships, and wherein dxdy indicates the physics size of each pixel
It is write above formula (3) as matrix form, is obtained
Formula (2) is brought into formula (4), is obtained
It puts aside translation problem, enables u0=v0=0,fx=fy=7.1254 × 102
Dividing vehicle area is S under pixel coordinate systempixel
It is S that dividing vehicle area is wherein corresponded under real spacereal=xy, Z axis distance correspond to image under real space Distance d of the head apart from front truck, obtains
The front truck absolute depth being partitioned into accurate can be acquired in the case where unobstructed according to formula (8), for vehicle The case where being blocked, obtained being blocked after the exposure mask value of part according to 3D CAD model, formula (8) is still applicable in；For vehicle The case where being truncated is corresponded on 3D CAD model by the missing of exposure mask in 2D image and determines its shared length and width in three dimensions High size, thus to the S for being truncated vehiclerealIt adjusts, formula (8) is recycled to carry out absolute depth estimation.
Monocular vision is combined the absolute depth values of estimation vehicle by the present invention with image segmentation phase with objective detection, first The extraction for first carrying out two-dimentional bounding box and three-dimensional boundaries frame to target vehicle based on deep learning algorithm, obtains its corresponding position letter Breath；Secondly, the length, width and height based on three-dimensional boundaries frame are matched with 3D CAD auto model, the corresponding approximate three-dimensional vehicle of vehicle is obtained Model；Again and, two-dimentional bounding box is based on to vehicle extraction vehicle classification information in figure；Therewith, by the corresponding three-dimensional information of vehicle It is sent into example segmentation network with vehicle model information, according to camera imaging principle, calculates image according to the dimension information of different automobile types The absolute depth values of middle vehicle.The present invention meets the requirement of front truck real time distance under vehicle-mounted camera, only carries out to vehicle absolute Depth detection, while solving the problems, such as overlapping or truncated region between vehicle, so that the vehicle for being blocked or being truncated still can be with Accurate its absolute depth values of detection.
The above are preferred embodiments of the present invention, all any changes made according to the technical solution of the present invention, and generated function is made When with range without departing from technical solution of the present invention, all belong to the scope of protection of the present invention.

Claims (4)

1. the front truck distance measuring method under a kind of vehicle-mounted camera based on monocular vision and image segmentation, which is characterized in that including with Lower step:
Step S1: image is read by frame from the video flowing that vehicle-mounted camera is shot；
Step S2: target detection is carried out to vehicle, extracts the location information of vehicle in image, including two-dimentional bounding box information and three Tie up bounding box information；
Step S3: the CAD model of corresponding vehicle is matched in preset vehicle 3D CAD model library according to three-dimensional boundaries frame information；
Step S4: the distance of overlapping vehicle is judged according to the two-dimentional bounding box information of vehicle, according to a left side for vehicle two dimension bounding box Upper angular vertex ordinate value, the exposure mask of CAD model projection on the image determine the block information of vehicle；
Step S5: vehicle is detected according to the two-dimentional bounding box information of vehicle, the dimension information of vehicle is obtained according to specific vehicle；Institute The dimension information for stating vehicle is the length, width and height of vehicle；
Step S6: corresponding position vehicle in image is obtained using example segmentation network according to the dimension information of vehicle and block information Pixel-level exposure mask value；
Step S7: according to camera imaging principle, it is corresponding that each car is calculated using the dimension information of vehicle and the result of vehicles segmentation Absolute depth values and the Some vehicles absolute depth values that are blocked；
The step S7 is specifically included:
According to camera imaging principle, fathomed using the relationship between the elemental area of vehicle and true area, in image In coordinate system, the coordinate of the two-dimentional bounding box top left corner apex of vehicle is q (x, y, f), in actual spatial coordinates system, vehicle The coordinate of two-dimentional bounding box top left corner apex is Q (X, Y, Z), then
It is write as matrix form, is obtained
Pixel coordinate system is defined as (u, v), Q1(u0,v0) be image coordinate system plane and optical axis intersection point, in image coordinate system Point coordinate be (x, y), be reflected as in pixel coordinate system (u, v), then image coordinate system and pixel coordinate system exist with ShiShimonoseki System, wherein dxdy indicates the physics size of each pixel
It is write above formula (3) as matrix form, is obtained
Formula (2) is brought into formula (4), is obtained
Do not consider translation problem, enables u0=v0=0,fx=fy=7.1254 × 102
Dividing vehicle area is S under pixel coordinate systempixel
It is S that dividing vehicle area is wherein corresponded under real spacereal=xy, Z axis coordinate value correspond under real space camera away from From front truck distance d, obtain
2. front truck distance measuring method according to claim 1, which is characterized in that the step S3 is specifically included: according to three-dimensional Bounding box information is chosen closest at high proportion with the length and width of the cuboid of target vehicle in preset vehicle 3D CAD model library Model, by the 3D CAD model mapping be matched in the 3D cuboid of respective objects vehicle.
3. front truck distance measuring method according to claim 1, which is characterized in that the step S4 is specifically included: being sat in image In mark system, the top left corner apex ordinate value of the two-dimentional bounding box of vehicle is compared, the small two-dimentional bounding box of ordinate value is corresponding Vehicle is nearby vehicle, and the big corresponding vehicle of two-dimentional bounding box of ordinate value is remote vehicle, according to vehicle two dimension bounding box Top left corner apex ordinate value, CAD model projection exposure mask on the image determines the visible part of target vehicle, is blocked Part and the part being truncated by image.
4. front truck distance measuring method according to claim 1, which is characterized in that the vehicle includes multi-functional Recreational Vehicle, fortune Ejector half vehicle, wing-rooms on either side of a one-story house type hatchback vehicle, four-door sedan, mini bus, streamlined car, four three compartment extended type station wagons, Picking-up vehicle, transboundary Recreational Vehicle, open car, sport car and steel plate top sport car.
CN201810529462.0A 2018-05-29 2018-05-29 Front truck distance measuring method under vehicle-mounted camera based on monocular vision and image segmentation Expired - Fee Related CN108759667B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810529462.0A CN108759667B (en) 2018-05-29 2018-05-29 Front truck distance measuring method under vehicle-mounted camera based on monocular vision and image segmentation

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810529462.0A CN108759667B (en) 2018-05-29 2018-05-29 Front truck distance measuring method under vehicle-mounted camera based on monocular vision and image segmentation

Publications (2)

Publication Number Publication Date
CN108759667A CN108759667A (en) 2018-11-06
CN108759667B true CN108759667B (en) 2019-11-12

Family

ID=64003149

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810529462.0A Expired - Fee Related CN108759667B (en) 2018-05-29 2018-05-29 Front truck distance measuring method under vehicle-mounted camera based on monocular vision and image segmentation

Country Status (1)

Country Link
CN (1) CN108759667B (en)

Families Citing this family (31)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109658419B (en) * 2018-11-15 2020-06-19 浙江大学 Method for segmenting small organs in medical image
CN111256651B (en) * 2018-12-03 2022-06-07 北京京东乾石科技有限公司 Week vehicle distance measuring method and device based on monocular vehicle-mounted camera
CN111386530A (en) * 2018-12-29 2020-07-07 深圳市大疆创新科技有限公司 Vehicle detection method and apparatus
CN109726692A (en) * 2018-12-29 2019-05-07 重庆集诚汽车电子有限责任公司 High-definition camera 3D object detection system based on deep learning
CN109859489A (en) * 2019-01-30 2019-06-07 驭势科技（北京）有限公司 A kind of spacing estimation method, device, mobile unit and storage medium
CN109785661A (en) * 2019-02-01 2019-05-21 广东工业大学 A kind of parking guide method based on machine learning
KR102190527B1 (en) * 2019-02-28 2020-12-14 현대모비스 주식회사 Apparatus and method for automatic synthesizing images
CN109919072B (en) * 2019-02-28 2021-03-19 桂林电子科技大学 Fine vehicle type recognition and flow statistics method based on deep learning and trajectory tracking
US11276189B2 (en) * 2019-03-06 2022-03-15 Qualcomm Incorporated Radar-aided single image three-dimensional depth reconstruction
CN109902643B (en) * 2019-03-07 2021-03-16 浙江啄云智能科技有限公司 Intelligent security inspection method, device and system based on deep learning and electronic equipment thereof
CN109917359B (en) * 2019-03-19 2022-10-14 福州大学 Robust vehicle distance estimation method based on vehicle-mounted monocular vision
CN110060298B (en) * 2019-03-21 2023-06-20 径卫视觉科技（上海）有限公司 Image-based vehicle position and posture determining system and corresponding method
CN110297232A (en) * 2019-05-24 2019-10-01 合刃科技（深圳）有限公司 Monocular distance measuring method, device and electronic equipment based on computer vision
CN110307791B (en) * 2019-06-13 2020-12-29 东南大学 Vehicle length and speed calculation method based on three-dimensional vehicle boundary frame
CN110517349A (en) * 2019-07-26 2019-11-29 电子科技大学 A kind of 3D vehicle target detection method based on monocular vision and geometrical constraint
CN112580402A (en) * 2019-09-30 2021-03-30 广州汽车集团股份有限公司 Monocular vision pedestrian distance measurement method and system, vehicle and medium thereof
CN110992304B (en) * 2019-10-30 2023-07-07 浙江力邦合信智能制动***股份有限公司 Two-dimensional image depth measurement method and application thereof in vehicle safety monitoring
CN111009166B (en) * 2019-12-04 2021-06-01 上海市城市建设设计研究总院（集团）有限公司 Road three-dimensional sight distance checking calculation method based on BIM and driving simulator
CN111126248A (en) * 2019-12-20 2020-05-08 湖南千视通信息科技有限公司 Method and device for identifying shielded vehicle
CN111046843B (en) * 2019-12-27 2023-06-20 华南理工大学 Monocular ranging method in intelligent driving environment
CN111351436B (en) * 2020-03-06 2021-06-18 大连理工大学 Method for verifying precision of structural plane displacement vision measurement system
CN112036389B (en) * 2020-11-09 2021-02-02 天津天瞳威势电子科技有限公司 Vehicle three-dimensional information detection method, device and equipment and readable storage medium
CN112816496B (en) * 2021-01-05 2022-09-23 广州市华颉电子科技有限公司 Automatic optical detection method and device for interface assembly quality of automobile domain controller
CN113225447B (en) * 2021-04-14 2024-03-26 思看科技(杭州)股份有限公司 Three-dimensional scanning system, data processing method, data processing device and computer equipment
CN113221739B (en) * 2021-05-12 2023-04-14 中国科学技术大学 Monocular vision-based vehicle distance measuring method
CN113269118B (en) * 2021-06-07 2022-10-11 重庆大学 Monocular vision forward vehicle distance detection method based on depth estimation
CN114049394B (en) * 2021-11-23 2022-06-21 智道网联科技(北京)有限公司 Monocular distance measuring method, device, equipment and storage medium
CN114882727B (en) * 2022-03-15 2023-09-05 深圳市德驰微视技术有限公司 Parking space detection method based on domain controller, electronic equipment and storage medium
CN116012453B (en) * 2023-03-28 2023-06-09 常州星宇车灯股份有限公司 Monocular distance measuring method, device, equipment and medium for intelligent driving of vehicle and vehicle
CN116500796A (en) * 2023-06-30 2023-07-28 合肥疆程技术有限公司 Projection processing method, head-up display, automobile and storage medium
CN117854055A (en) * 2024-03-07 2024-04-09 河南百合特种光学研究院有限公司 Front vehicle distance judging method for license plate recognition and monocular vision of vehicle-mounted camera

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP3428876A1 (en) * 2016-03-10 2019-01-16 Ricoh Company, Ltd. Image processing device, apparatus control system, imaging device, image processing method, and program

Family Cites Families (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101408978B (en) * 2008-11-27 2010-12-01 东软集团股份有限公司 Method and apparatus for detecting barrier based on monocular vision
CN103287372B (en) * 2013-06-19 2015-09-23 贺亮才 A kind of automobile collision preventing method for security protection based on image procossing
CN104899554A (en) * 2015-05-07 2015-09-09 东北大学 Vehicle ranging method based on monocular vision
JPWO2018030319A1 (en) * 2016-08-12 2018-08-09 パナソニックＩｐマネジメント株式会社 Ranging system and mobile system
WO2018042954A1 (en) * 2016-08-29 2018-03-08 日立オートモティブシステムズ株式会社 On-vehicle camera, method for adjusting on-vehicle camera, and on-vehicle camera system
CN106826815B (en) * 2016-12-21 2019-05-31 江苏物联网研究发展中心 The method with positioning is identified based on the target object of color image and depth image
CN107390205B (en) * 2017-07-20 2019-08-09 清华大学 A kind of monocular vision vehicle odometry method obtaining front truck feature using car networking
CN107796373B (en) * 2017-10-09 2020-07-28 长安大学 Distance measurement method based on monocular vision of front vehicle driven by lane plane geometric model
CN107972662B (en) * 2017-10-16 2019-12-10 华南理工大学 Vehicle forward collision early warning method based on deep learning

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP3428876A1 (en) * 2016-03-10 2019-01-16 Ricoh Company, Ltd. Image processing device, apparatus control system, imaging device, image processing method, and program

Also Published As

Publication number Publication date
CN108759667A (en) 2018-11-06

Similar Documents

Publication Publication Date Title
CN108759667B (en) Front truck distance measuring method under vehicle-mounted camera based on monocular vision and image segmentation
CN111126269B (en) Three-dimensional target detection method, device and storage medium
CN110065494B (en) Vehicle anti-collision method based on wheel detection
CN105300403B (en) A kind of vehicle mileage calculating method based on binocular vision
CN110745140B (en) Vehicle lane change early warning method based on continuous image constraint pose estimation
CN108638999B (en) Anti-collision early warning system and method based on 360-degree look-around input
CN104933409B (en) A kind of parking stall recognition methods based on panoramic picture dotted line feature
CN102774325B (en) Rearview reversing auxiliary system and method for forming rearview obstacle images
CN109359409A (en) A kind of vehicle passability detection system of view-based access control model and laser radar sensor
CN109631896A (en) A kind of parking lot autonomous parking localization method based on vehicle vision and motion information
CN110517349A (en) A kind of 3D vehicle target detection method based on monocular vision and geometrical constraint
US11887336B2 (en) Method for estimating a relative position of an object in the surroundings of a vehicle and electronic control unit for a vehicle and vehicle
CN103020948A (en) Night image characteristic extraction method in intelligent vehicle-mounted anti-collision pre-warning system
CN110619674B (en) Three-dimensional augmented reality equipment and method for accident and alarm scene restoration
CN108106627A (en) A kind of monocular vision vehicle positioning method of the online dynamic calibration of distinguished point based
CN110969064A (en) Image detection method and device based on monocular vision and storage equipment
CN110060298A (en) A kind of vehicle location and attitude and heading reference system based on image and corresponding method
CN108645375A (en) One kind being used for vehicle-mounted biocular systems rapid vehicle distance measurement optimization method
CN115082881A (en) Target detection method, storage medium, electronic device, and vehicle
CN114359744A (en) Depth estimation method based on fusion of laser radar and event camera
CN114291011A (en) Vehicle neural network
CN116978009A (en) Dynamic object filtering method based on 4D millimeter wave radar
Gruyer et al. Vehicle detection and tracking by collaborative fusion between laser scanner and camera
Omar et al. Detection and localization of traffic lights using YOLOv3 and Stereo Vision
Jung et al. Intelligent Hybrid Fusion Algorithm with Vision Patterns for Generation of Precise Digital Road Maps in Self-driving Vehicles.

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
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20191112

CF01 Termination of patent right due to non-payment of annual fee