CN109920001A - Method for estimating distance based on pedestrian head height - Google Patents
Method for estimating distance based on pedestrian head height Download PDFInfo
- Publication number
- CN109920001A CN109920001A CN201910194211.6A CN201910194211A CN109920001A CN 109920001 A CN109920001 A CN 109920001A CN 201910194211 A CN201910194211 A CN 201910194211A CN 109920001 A CN109920001 A CN 109920001A
- Authority
- CN
- China
- Prior art keywords
- distance
- head
- pedestrian
- height
- interpolation
- 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
Landscapes
- Traffic Control Systems (AREA)
- Length Measuring Devices By Optical Means (AREA)
- Image Analysis (AREA)
Abstract
Method for estimating distance based on pedestrian head height, belong to intelligent image visible sensation distance estimation field, for solving the reunification of the high performance-price ratio in people's vehicle distance, hard real-time and low complex degree, main points are to obtain different distance top component pixels tall in the picture；Rigid element size-is constructed apart from interpolation model；Obtain pedestrian head part pixels tall to be measured in image, and it is inputted as rigid element size-apart from interpolation model, distance is obtained, effect is that rigid element size-is small apart from interpolation model calculation amount, can satisfy requirement of the Autonomous Vehicle for distance estimations system in terms of real-time.
Description
Technical field
The invention belongs to intelligent image visible sensation distances to estimate field, specifically a kind of to pass through pedestrian head pixels tall
Come the method estimated of adjusting the distance.
Background technique
With popularizing for automobile, vehicle gradually obtains the extensive concern of people with pedestrains safety problem.As pedestrian's danger
The main foundation determined, various distance estimations systems are also come into being.Distance estimations system can not only carry out pedestrian's danger
Judgement, while data support can be provided for vehicle track planning, pedestrian's avoidance etc., become the hot spot of Recent study.
It is many kinds of for the method for distance now, it can specifically be divided into: be adjusted the distance and estimated based on visible light,
It is adjusted the distance based on infrared light and estimate and adjust the distance by a variety of distance mearuring equipments being estimated.
Algorithm is usually by increasing the complexity of requirement or algorithm to hardware come the essence of range of lift estimated result now
Degree.But the cost of system entirety is increased from increasing to start with to the angle of hardware requirement, it is difficult to be widely applied, and calculate from being promoted
Start in terms of method complexity and will seriously affect the real-time of distance estimating algorithm.Therefore, it is necessary to develop it is a kind of with it is low at
The monocular cam of this deployment is platform, and is capable of the method for quick obtaining distance, to meet Autonomous Vehicle DAS (Driver Assistant System)
It is required that.
In general, raising distance estimations result precision, which is generallyd use, merges by a variety of distance mearuring equipments or increases algorithm
The mode of complexity.But the above method is usually difficult to reach wanting for vehicle-mounted DAS (Driver Assistant System) in terms of popularization degree and real-time
It asks.Patent " a kind of road ahead pedestrian fast ranging method based on vehicle-mounted binocular camera " (publication number: CN104700414A)
It proposes and a kind of is adjusted the distance the method estimated by binocular scene.This method passes through the visual difference pair of left image and right image
Distance is estimated that the point in left and right two image obtained with color characteristic and scale invariant feature to binocular camera carries out
Matching solves the problems, such as that visual difference estimates inaccuracy using parallax intermediate value.Patent " the front pedestrian ranging side of view-based access control model
Method " (publication number: CN104966064A) by camera acquire 20 checkerboard grid images, and utilize MATLAB to calculate camera
Inner parameter comprehensively considers and calculates the camera external parameter obtained by acquisition image, proposes that image coordinate system and the world are sat
Mapping relations between mark system, to carry out the estimation of distance.
Summary of the invention
In order to solve the reunification of the high performance-price ratio in people's vehicle distance, hard real-time and low complex degree, the present invention is proposed
Following technical solution:
A kind of method for estimating distance based on height of head, comprising:
Step 1: different distance top component pixels tall is obtained in the picture；
Step 2: building rigid element size-is apart from interpolation model；
Step 3: pedestrian head part pixels tall to be measured in image is obtained, and as rigid element size-apart from interpolation mould
Type input, obtains distance.
Further, specific step is as follows for step step 2: first to multiple target pedestrian head pictures in different distance
Plain height is obtained, and the n pedestrian head pixels tall that enabling is fixed value d from camera distance but position is different is pi, right
This carries out equalization processing apart from upper head pixels tall, acquires interpolation height node
Height of head-distance relationship is sought according to Lagrange interpolation, if n head height interpolation node is P, relatively
The corresponding relationship of the landmark point measured value D for the one group of pedestrian target distance answered, height of head and distance is to ask an interpolation more
Item formula H, meets relationship,
Wherein,Above-mentioned formula acquirement result is unique, then obtains rigid portion
Part size-is apart from interpolation model.
Further, specific step is as follows for step 3: obtaining the image of the pedestrian containing target by monocular cam, obtains
Target pedestrian head pixels tall q' inputs rigid element size-in interpolation model, obtains target pedestrian and camera
The distance between.
The utility model has the advantages that
The present invention is based on intelligent image, the altitude response shown in different distance using head, in conjunction with right
The depth of image information is excavated, and realizes distance estimations.Present invention is mainly applied in DAS (Driver Assistant System), hardware aspect only
The cost of system is reduced by the design of monocular cam, it is easy to spread；By the head pixel for measuring multiple target pedestrians
Height obtains interpolation knot, and obtains the corresponding relationship between rigid element size and distance by Lagrange interpolation method.By
In in same distance, different pedestrians can express identical head pixels tall, therefore the rigid element size-is apart from interpolation mould
Type has very strong universality, reduces subsequent redundant computation；In actual measurement, target pedestrian head is obtained by image
After pixels tall, input rigid element size-can obtain distance in interpolation model.Rigid element size-is apart from interpolation
Model calculation amount is small, can satisfy requirement of the Autonomous Vehicle for distance estimations system in terms of real-time.Cost performance of the present invention
High, strong real-time has certain superiority in terms of Autonomous Vehicle assists driving, and promotional value is high.
The present invention constructs rigid element size-apart from interpolation model, by repeatedly to targets multiple in different distance first
Pedestrian head pixels tall is sampled, using mean value as to the same standardized hand of height of head apart from upper different pedestrians
Section.The actual distance between pedestrian and camera is combined after acquisition interpolation knot, establishes rigid element size-distance model.When
When image inputs again, it is input in model after obtaining pedestrian head pixels tall, can be obtained target pedestrian and camera
The distance between.
The present invention can estimate distance between vehicle and pedestrian, suitable for Autonomous Vehicle auxiliary driving for row
People's danger anticipation provides effective data supporting, and algorithm process method is simple, and exploitativeness is strong；Using rigid element size-
Constructed apart from interpolation model, input two step of head pixels tall is adjusted the distance and is estimated, model once establish, without it is secondary more
Change；Distance estimations only can be completed using monocular cam as image capture device, it is low to hardware-dependent degree, it is easy to accomplish.
Detailed description of the invention
Fig. 1 is using the distance estimating algorithm principle logic chart based on height of head；
Fig. 2 is that rigid element height obtains result；
Fig. 3 is the testing result figure of embodiment 1；
Fig. 4 is the testing result figure of embodiment 2；
Specific embodiment
Present invention is further described in detail with specific embodiment with reference to the accompanying drawing:
A kind of principle logic chart of the method for estimating distance based on height of head is as shown in Figure 1, algorithm specific implementation step
It is rapid as follows:
Step 1: different distance top component pixels tall is obtained by existing algorithm；
Step 2: building rigid element size-is apart from interpolation model；
Step 3: pedestrian head part pixels tall to be measured in image is obtained, and as rigid element size-apart from interpolation mould
Type input, obtains distance；
Specific embodiments of the present invention are as follows:
Emphasis of the present invention is to be judged using the altitude response of head part the distance between target pedestrian and camera.
Head part is unlikely to deform as rigid element, and most of pedestrian head height is relatively uniform, in controlled range, utilizes head
The pixels tall difference that part is shown in different distance can carry out distance estimations.After obtaining image by monocular cam, root
Pedestrian head pixels tall is obtained according to existing algorithm.It is established using the head part altitude response that pedestrian shows in different distance
Rigid element size-shows height of head and distance relation apart from interpolation model.When image inputs again, calculated according to having
Method obtains head pixels tall, is input to rigid element size-in interpolation model, distance can be obtained.Principle logic
Figure is as shown in Figure 1.
With the development of auxiliary driving technology, all kinds of method for estimating distance also emerge one after another.In these method for estimating distance
In, the method for relatively conventional raising distance estimations result precision is the requirement promoted to hardware or the complexity for increasing algorithm.
But for mobile unit, higher requirement is suffered to cost performance and real-time aspect, existing distance measuring method is difficult full
Foot.Therefore, the method for estimating distance for developing a kind of high real-time low cost just seems outstanding with application for the exploitation of Autonomous Vehicle
It is important.In view of the above problems, the present invention devises a kind of method for estimating distance based on height of head.This method obtains first
The height of head feature that pedestrian is shown in different distance constructs rigid element size-apart from interpolation model.When having again
When image inputs, target pedestrian head height is input in model, to obtain distance value.This method is to hardware-dependent journey
Spend it is low, only using monocular cam as image acquisition equipment.In terms of real-time, rigid element size-apart from interpolation model only
It need to disposably construct, reduce the redundancy on calculating.
Furtherly, this method is by rigid element size-apart from interpolation model building, input height of head two parts group
At specific as follows:
(1) rigid element size-is apart from Quick estimator
Foundation of the rigid element size-apart from interpolation model determines final estimation as first link of the invention
The precision of distance.Its essence is the altitude response that is shown in different distance by rigid element head to camera and mesh
Distance between mark pedestrian judges.In model foundation, first to multiple target pedestrian head pixels in different distance
Height is obtained, and it is as shown in Figure 2 to obtain result.The n pedestrian head that enabling is fixed value d from camera distance but position is different
Pixels tall is pi, equalization processing is carried out apart from upper head pixels tall to this, acquires interpolation height node
Height of head-distance relationship is sought according to Lagrange interpolation.If n head height interpolation node is P, relatively
The landmark point measured value D for the one group of pedestrian target distance answered, the corresponding relationship of height of head and distance as ask an interpolation more
Item formula H, meets relationship,
Wherein,It is unique that above-mentioned formula obtains result.
2) distance obtains
The image of the pedestrian containing target is obtained by monocular cam, obtains target pedestrian head pixels tall q', and input is rigid
Property part dimension-can obtain the distance between target pedestrian and camera in interpolation model.
It is one kind by rigid element in image the invention proposes a kind of method for estimating distance based on height of head
The method that the acquisition of size carries out distance estimations.In characteristics of human body, head is rigid element, and different pedestrian head differences in height
Away from smaller.Accurately the distance between target pedestrian and camera can be estimated using height of head as feature.
For above-mentioned reason, the present invention shows altitude feature building by human body rigid element head in different distance
Rigid element size-is apart from interpolation model.When image inputs again, obtains head pixels tall and input rigid element ruler
Very little-in interpolation model, it can be obtained distance.The present invention is using model construction, input two step pair of head part pixels tall
Distance is estimated.
It can be found by consulting literatures, being adjusted the distance by target object size and carrying out judgement is in monocular cam telemetry
A kind of more common means.And for pedestrian's ranging, head is as rigid element and adult's height of head is 23cm
Left and right is relatively fixed, can be used as the standard of distance estimations by the altitude feature embodied in different distance.
Through emulation testing, the present invention only relies on monocular cam as image capture device in hardware aspect, cheap,
Generalization is extremely strong；Algorithm is relatively easy, can satisfy the requirement of assistant equipments on board in terms of real-time；In computational accuracy side
Face, can be by control errors within 0.5m within the scope of 10m, can be by control errors within 1m within the scope of 30m.The present invention exists
Assist driving field application high, it is practical, there is extremely strong promotional value.
Embodiment 1:
This embodiment example is completed experimental image using vehicle-mounted camera (3000*4000@30Hz) and is acquired, this image is in spring
It completes to shoot on street, pedestrian containing pedestrian target 3 in image, be taken away from the walking of camera direction, before camera respectively
The posture crossed.Distance between target pedestrian and camera is estimated through the invention, the two distance estimations result and true
Actual distance is from as shown in figure 3, error refers to no more than 1m.
Embodiment 2:
This embodiment example is completed experimental image using vehicle-mounted camera (3000*4000@30Hz) and is acquired, this image is in spring
It completes to shoot on street, pedestrian containing pedestrian target 5 in image, is respectively adopted and walks towards camera direction, cross camera
Front crosses zebra stripes, walks away from camera and plays the posture of phone in roadside.Through the invention to target pedestrian and camera shooting
Distance is estimated that the two distance estimations result and actual distance are as shown in Figure 4 between head.Error amount is no more than 1m.
Claims (3)
1. a kind of method for estimating distance based on height of head characterized by comprising
Step 1: different distance top component pixels tall is obtained in the picture；
Step 2: building rigid element size-is apart from interpolation model；
Step 3: pedestrian head part pixels tall to be measured in image is obtained, and defeated apart from interpolation model as rigid element size-
Enter, obtains distance.
2. as described in claim 1 based on the method for estimating distance of height of head, which is characterized in that step step 2 it is specific
Steps are as follows: obtaining first to multiple target pedestrian head pixels talls in different distance, order is with camera distance
The different n pedestrian head pixels tall of fixed value d but position is pi, this is equalized apart from upper head pixels tall
Processing, acquires interpolation height node
Height of head-distance relationship is sought according to Lagrange interpolation, it is corresponding if n head height interpolation node is P
The corresponding relationship of the landmark point measured value D of one group of pedestrian target distance, height of head and distance is to seek an interpolation polynomial
H meets relationship,
Wherein,Above-mentioned formula acquirement result is unique, then obtains rigid element ruler
Very little-apart from interpolation model.
3. as claimed in claim 2 based on the method for estimating distance of height of head, which is characterized in that the specific steps of step 3
It is as follows: to obtain the image of the pedestrian containing target by monocular cam, obtain target pedestrian head pixels tall q', input rigid portion
Part size-obtains the distance between target pedestrian and camera in interpolation model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910194211.6A CN109920001A (en) | 2019-03-14 | 2019-03-14 | Method for estimating distance based on pedestrian head height |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910194211.6A CN109920001A (en) | 2019-03-14 | 2019-03-14 | Method for estimating distance based on pedestrian head height |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109920001A true CN109920001A (en) | 2019-06-21 |
Family
ID=66964955
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910194211.6A Pending CN109920001A (en) | 2019-03-14 | 2019-03-14 | Method for estimating distance based on pedestrian head height |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109920001A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111126178A (en) * | 2019-12-05 | 2020-05-08 | 大连民族大学 | Infrared-visible light binocular pedestrian body multi-component fusion continuous distance estimation method |
| CN111145211A (en) * | 2019-12-05 | 2020-05-12 | 大连民族大学 | Monocular camera upright pedestrian head pixel height acquisition method |
| CN111738088A (en) * | 2020-05-25 | 2020-10-02 | 西安交通大学 | Pedestrian distance prediction method based on monocular camera |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104966064A (en) * | 2015-06-18 | 2015-10-07 | 奇瑞汽车股份有限公司 | Pedestrian ahead distance measurement method based on visual sense |
| CN106909929A (en) * | 2015-12-22 | 2017-06-30 | 比亚迪股份有限公司 | Pedestrian's distance detection method and device |
-
2019
- 2019-03-14 CN CN201910194211.6A patent/CN109920001A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104966064A (en) * | 2015-06-18 | 2015-10-07 | 奇瑞汽车股份有限公司 | Pedestrian ahead distance measurement method based on visual sense |
| CN106909929A (en) * | 2015-12-22 | 2017-06-30 | 比亚迪股份有限公司 | Pedestrian's distance detection method and device |
Non-Patent Citations (1)
| Title |
|---|
| 程凡等: "单目视觉行人身体多部件距离融合估计算法", 《大连民族大学学报（自然科学）》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111126178A (en) * | 2019-12-05 | 2020-05-08 | 大连民族大学 | Infrared-visible light binocular pedestrian body multi-component fusion continuous distance estimation method |
| CN111145211A (en) * | 2019-12-05 | 2020-05-12 | 大连民族大学 | Monocular camera upright pedestrian head pixel height acquisition method |
| CN111145211B (en) * | 2019-12-05 | 2023-06-30 | 大连民族大学 | Method for acquiring pixel height of head of upright pedestrian of monocular camera |
| CN111126178B (en) * | 2019-12-05 | 2023-07-04 | 大连民族大学 | Continuous distance estimation method for infrared-visible light binocular pedestrian body multi-component fusion |
| CN111738088A (en) * | 2020-05-25 | 2020-10-02 | 西安交通大学 | Pedestrian distance prediction method based on monocular camera |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10937231B2 (en) | Systems and methods for updating a high-resolution map based on binocular images | |
| CN107389026B (en) | A kind of monocular vision distance measuring method based on fixed point projective transformation | |
| CN111436216B (en) | Method and system for color point cloud generation | |
| US11433880B2 (en) | In-vehicle processing apparatus | |
| CN106407315B (en) | A kind of vehicle autonomic positioning method based on street view image database | |
| CN107886477B (en) | Fusion correction method for three-dimensional vision and low-beam laser radar in unmanned driving | |
| CN110859044B (en) | Integrated sensor calibration in natural scenes | |
| CN107390205B (en) | A kind of monocular vision vehicle odometry method obtaining front truck feature using car networking | |
| JP2022110001A (en) | Method and system for generating and using localization reference data | |
| CN109920001A (en) | Method for estimating distance based on pedestrian head height | |
| JP2020528994A (en) | Vehicle navigation system using attitude estimation based on point cloud | |
| CN110567469A (en) | Visual positioning method and device, electronic equipment and system | |
| KR102226846B1 (en) | System for Positioning Hybrid Indoor Localization Using Inertia Measurement Unit Sensor and Camera | |
| WO2020140164A1 (en) | Systems and methods for updating a high-definition map | |
| JP6779365B2 (en) | Object detection device and vehicle | |
| CN103308058A (en) | Enhanced data association of fusion using weighted bayesian filtering | |
| CN104281148A (en) | Mobile robot autonomous navigation method based on binocular stereoscopic vision | |
| Nienaber et al. | A comparison of low-cost monocular vision techniques for pothole distance estimation | |
| JP6838285B2 (en) | Lane marker recognition device, own vehicle position estimation device | |
| CN110793544A (en) | Sensing sensor parameter calibration method, device, equipment and storage medium | |
| CN111510704B (en) | Method for correcting camera dislocation and device using same | |
| CN114719848B (en) | Unmanned aerial vehicle height estimation method based on vision and inertial navigation information fusion neural network | |
| CN113740871B (en) | Laser SLAM method, system equipment and storage medium under high dynamic environment | |
| JP2017181476A (en) | Vehicle location detection device, vehicle location detection method and vehicle location detection-purpose computer program | |
| CN112017238A (en) | Method and device for determining spatial position information of linear object |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190621 |