CN109584297A - Object detection method and device - Google Patents
Object detection method and device Download PDFInfo
- Publication number
- CN109584297A CN109584297A CN201811242524.6A CN201811242524A CN109584297A CN 109584297 A CN109584297 A CN 109584297A CN 201811242524 A CN201811242524 A CN 201811242524A CN 109584297 A CN109584297 A CN 109584297A
- Authority
- CN
- China
- Prior art keywords
- image
- depth image
- depth
- testing result
- detection target
- 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.)
- Withdrawn
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- 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
- G06T7/55—Depth or shape recovery from multiple images
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
- G06T2207/10021—Stereoscopic video; Stereoscopic image sequence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10028—Range image; Depth image; 3D point clouds
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Geometry (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of object detection method and devices, are related to field of computer technology.This method comprises: obtaining color image；Obtain depth image；According to color image, detection target is obtained；According to depth image and detection target, testing result is obtained.In the present invention, by obtaining depth image, and color image is combined, detection accuracy can not only be improved, and can be improved detection speed.
Description
Technical field
The present invention relates to field of computer technology, in particular to object detection method and device.
Background technique
Carrying out target detection using traditional camera is a key areas in computation vision, is many computation vision perception
The basic algorithm in the world.Target detection origin algorithm is stencil matching, is slided from left to right, from top to bottom using sliding window,
Classified and identifies target.Later, convolutional neural networks (Convolutional Neural Network, CNN) instead of
Original sliding window mode, becomes the mainstream thoughts of sorting algorithm.In the prior art, by carrying out candidate regions (Region of
Interest, ROI) selection, carry out CNN classification, by using the higher ROI of less and quality, improve sliding window speed and
Precision, this method need very more candidate regions to promote accuracy rate.
But the prior art has many regions to overlap each other in the candidate regions of selection, therefore this method is to target detection
Precision it is lower.
Summary of the invention
The embodiment of the invention provides a kind of object detection method and devices.It aims to solve the problem that target detection precision is lower to ask
Topic.In order to which some aspects of the embodiment to disclosure have a basic understanding, simple summary is shown below.The summary portion
Dividing is not extensive overview, nor to determine key/critical component or describe the protection scope of these embodiments.It is unique
Purpose is that some concepts are presented with simple form, in this, as the preamble of following detailed description.
According to a first aspect of the embodiments of the present invention, a kind of object detection method is provided, comprising:
Obtain color image；
Obtain depth image；
According to color image, detection target is obtained；
According to depth image and detection target, testing result is obtained.
Optionally, after obtaining detection target, further includes:
According to detection target, candidate regions are obtained；
According to depth image and detection target, testing result is obtained, comprising:
According to depth image and candidate regions, testing result is obtained.
Optionally, according to depth image and candidate regions, testing result is obtained, comprising:
According to depth image, mask figure is obtained；
According to mask figure and candidate regions, testing result is obtained.
Optionally, according to depth image, mask figure is obtained, comprising:
The different mask number of the element marking that depth is greater than depth threshold on depth image；
According to mask number, mask figure is obtained.
Optionally, testing result includes detecting classification and the position of target.
Optionally, further includes:
According to color image and depth image, binocular vision matching image is obtained.
Optionally, according to color image and depth image, binocular vision matching image is obtained, comprising:
Internal reference matrix A 1, the internal reference matrix A 2 of depth image and the color image of color image are obtained to the rotation of depth image
Torque battle array R, translation matrix T；
According to A1, A2, R and T, binocular vision matching image is obtained.
According to a second aspect of the embodiments of the present invention, a kind of object detecting device is provided, comprising:
Colour imagery shot, for obtaining color image；
Depth camera, for obtaining depth image；
Processor, the color image for being obtained according to colour imagery shot obtain detection target；
The depth image and detection target obtained according to depth camera, obtains testing result.
Technical solution disclosed by the invention, the position by obtaining depth image, between available object and camera
Relationship can further obtain the physical location and size of object in conjunction with color image.It, can after the actual size for restoring object
So that anchor point position is relatively fixed, and reduces the quantity of anchor point, detection accuracy can not only be improved, and can be improved detection speed
Degree.
It should be understood that above general description and following detailed description be only it is exemplary and explanatory, not
It can the limitation present invention.
Detailed description of the invention
The drawings herein are incorporated into the specification and forms part of this specification, and shows and meets implementation of the invention
Example, and be used to explain the principle of the present invention together with specification.
Fig. 1 is a kind of flow chart of object detection method disclosed by the embodiments of the present invention；
Fig. 2 is a kind of schematic diagram of anchor point disclosed by the embodiments of the present invention；
Fig. 3 is a kind of schematic diagram of object detecting device disclosed by the embodiments of the present invention.
Specific embodiment
The following description and drawings fully show specific embodiments of the present invention, to enable those skilled in the art to
Practice them.Embodiment only represents possible variation.Unless explicitly requested, otherwise individual components and functionality is optional, and
And the sequence of operation can change.The part of some embodiments and feature can be included in or replace other embodiments
Part and feature.The range of embodiment of the present invention includes the entire scope of claims and the institute of claims
There is obtainable equivalent.Herein, each embodiment can individually or generally be indicated that this is only with term " invention "
It is merely for convenience, and if in fact disclosing the invention more than one, it is not meant to automatically limit the range of the application
For any single invention or inventive concept.Herein, relational terms such as first and second and the like are used only for one
Entity, which is perhaps operated, to be distinguished and exists without requiring or implying between these entities or operation with another entity or operation
Any actual relationship or sequence.Moreover, the terms "include", "comprise" or its any other variant be intended to it is non-exclusive
Property include so that include a series of elements process, method or equipment not only include those elements, but also including
Other elements that are not explicitly listed.Each embodiment herein is described in a progressive manner, and each embodiment stresses
Be the difference from other embodiments, the same or similar parts in each embodiment may refer to each other.For implementing
For structure, product etc. disclosed in example, since it is corresponding with part disclosed in embodiment, so being described relatively simple, phase
Place is closed referring to method part illustration.
The embodiment of the invention discloses a kind of object detection methods, as shown in Figure 1, comprising:
S101, color image is obtained；
S102, depth image is obtained；
S103, according to color image, obtain detection target；
S104, according to depth image and detection target, obtain testing result.
Wherein, in S101 and S102, color image and depth image are obtained respectively, and illustratively, color image can be with
It is obtained by colour imagery shot, such as common RGB camera, depth image can be obtained by depth camera, including but not limited to
Flight time (Time of Flight, TOF) imaging camera head, binocular imaging camera, the camera shooting of structure light video camera head even depth
Head, relative distance precision can achieve Centimeter Level.In S104, testing result may include detecting classification and the position of target
It sets.
In S103, as shown in Fig. 2, k anchor point, each anchor point can be arranged on each pixel of color image
A corresponding specific position, that is, need the k anchor point centered on each position, k prediction is carried out to the position.Particularly,
Each prediction is associated with specific anchor, but different location can share the anchor point of same shape.Those skilled in the art can be with
According to actual needs, shape and the position of anchor point are set, to cover the real-world objects with different proportion and the ratio of width to height.
After being imaged in video camera due to object, the information such as size and length-width ratio all can by the object and camera it
Between distance relation and positional relationship etc. influence.Technical solution disclosed by the invention, by obtaining depth image, available object
Positional relationship between body and camera can further obtain the physical location and size of object in conjunction with color image.Reduction
After the actual size of object, anchor point position can be made relatively fixed, and reduce the quantity of anchor point, detection essence can not only be improved
Degree, and can be improved detection speed.
Optionally, after S103, can also include:
S105, according to detection target, obtain candidate regions；
Then S104 may include:
According to depth image and candidate regions, testing result is obtained.
In the prior art, the positive rate of vacation that the repeatability of multiple candidate regions will cause target detection increases, object itself it
Between without overlapping or the lower situation of degree of overlapping under effect it is fine, but if in image there are when overlapped object if effect not
Ideal will cause great missing inspection situation.In technical solution disclosed by the invention, by obtaining and utilizing depth image can be very
Good solves the problems, such as this.When different objects has overlapping on traditional camera, can accurately be determined in depth image
The difference of overlapped object, to avoid the missing inspection situation as caused by being overlapped.
Optionally, S104 can also include:
S1041, according to depth image, obtain mask figure；
S1042, according to mask figure and candidate regions, obtain testing result.
Optionally, obtaining mask figure may include: that depth is greater than the element marking of depth threshold not on depth image
Same mask number obtains mask figure according to mask number.
Illustratively, the position of two pixels is respectively 1 and 2, and depth is respectively h1 and h2, when determining depth threshold
When for 0.2m, if | h1-h2 | > 0.2m, 1 and 2 two position will be labeled different mask numbers.It further, can be with
Zone marker by mask numbers different in mask figure is different color, and such as four color mask figures are carried out using 0,1,2,3 four number
The location of pixels that range information is greater than given threshold range is encoded into different mask numbers, overlapped each other for drawing by coding
Object boundary.
Optionally, this method can also include:
S106, according to color image and depth image, obtain binocular vision matching image.
Optionally, in S105, the internal reference matrix A 1 of available color image, the internal reference matrix A 2 of depth image and coloured silk
Spin matrix R, translation matrix T of the chromatic graph picture to depth image obtain binocular vision matching image according to A1, A2, R and T.
It is further alternative, binocular vision matching image can be obtained by Binocular Stereo Matching Algorithm.
The embodiment of the invention also discloses a kind of object detecting devices, as shown in Figure 3, comprising:
Colour imagery shot 301, for obtaining color image；
Depth camera 302, for obtaining depth image；
Processor 303, the color image for being obtained according to colour imagery shot 301 obtain detection target；It is taken the photograph according to depth
The depth image and detection target obtained as head, obtains testing result.
Optionally, testing result includes detecting classification and the position of target.
Optionally, processor 303 can be also used for:
According to detection target, candidate regions are obtained；
According to depth image and candidate regions, testing result is obtained.
Optionally, processor 303 can be also used for:
According to depth image, mask figure is obtained；
According to mask figure and candidate regions, testing result is obtained.
Optionally, processor 303 can be also used for:
The different mask number of the element marking that depth is greater than depth threshold on depth image；
According to mask number, mask figure is obtained.
Optionally, processor 303 can be also used for:
According to color image and depth image, binocular vision matching image is obtained.
Optionally, processor 303 can be also used for:
Internal reference matrix A 1, the internal reference matrix A 2 of depth image and the color image of color image are obtained to the rotation of depth image
Torque battle array R, translation matrix T；
According to A1, A2, R and T, binocular vision matching image is obtained.
It above are only example, those skilled in the art can also be combined into more in the case where not paying creative make great efforts
Mostly optional embodiment.
It should be understood that the invention is not limited to the process and structure that are described above and are shown in the accompanying drawings,
And various modifications and changes may be made without departing from the scope thereof.The scope of the present invention is only limited by the attached claims
System.
Claims (8)
1. a kind of object detection method characterized by comprising
Obtain color image；
Obtain depth image；
According to the color image, detection target is obtained；
According to the depth image and the detection target, testing result is obtained.
2. the method according to claim 1, wherein after obtaining the detection target, further includes:
According to the detection target, candidate regions are obtained；
According to the depth image and the detection target, the testing result is obtained, comprising:
According to the depth image and the candidate regions, the testing result is obtained.
3. according to the method described in claim 2, it is characterized in that, obtaining institute according to the depth image and the candidate regions
State testing result, comprising:
According to the depth image, mask figure is obtained；
According to the mask figure and the candidate regions, the testing result is obtained.
4. according to the method described in claim 3, wrapping it is characterized in that, obtain the mask figure according to the depth image
It includes:
The different mask number of the element marking that depth is greater than depth threshold on the depth image；
According to the mask number, the mask figure is obtained.
5. the method according to claim 1, wherein the testing result include it is described detection target classification and
Position.
6. the method according to claim 1, wherein further include:
According to the color image and the depth image, binocular vision matching image is obtained.
7. according to the method described in claim 4, it is characterized in that, being obtained according to the color image and the depth image
The binocular vision matching image, comprising:
The internal reference matrix A 1 of the color image, the internal reference matrix A 2 of the depth image and the color image are obtained described in
Spin matrix R, the translation matrix T of depth image；
According to described A1, A2, R and T, the binocular vision matching image is obtained.
8. a kind of object detecting device, comprising:
Colour imagery shot, for obtaining color image；
Depth camera, for obtaining depth image；
Processor, the color image for being obtained according to the colour imagery shot obtain detection target；
The depth image obtained according to the depth camera and the detection target obtain testing result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811242524.6A CN109584297A (en) | 2018-10-24 | 2018-10-24 | Object detection method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811242524.6A CN109584297A (en) | 2018-10-24 | 2018-10-24 | Object detection method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109584297A true CN109584297A (en) | 2019-04-05 |
Family
ID=65920486
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811242524.6A Withdrawn CN109584297A (en) | 2018-10-24 | 2018-10-24 | Object detection method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109584297A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109615647A (en) * | 2018-10-24 | 2019-04-12 | 北京升哲科技有限公司 | Object detection method and device |
| CN110207951A (en) * | 2019-05-23 | 2019-09-06 | 北京航空航天大学 | A kind of aircraft cable support assembled state detection method of view-based access control model |
| CN110220456A (en) * | 2019-06-26 | 2019-09-10 | 浙江大学 | A kind of hand-held box sizes measuring device and measuring method |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102509074A (en) * | 2011-10-18 | 2012-06-20 | Tcl集团股份有限公司 | Target identification method and device |
| US20160364912A1 (en) * | 2015-06-15 | 2016-12-15 | Electronics And Telecommunications Research Institute | Augmented reality-based hand interaction apparatus and method using image information |
| CN106778614A (en) * | 2016-12-16 | 2017-05-31 | 中新智擎有限公司 | A kind of human body recognition method and device |
| CN107368810A (en) * | 2017-07-20 | 2017-11-21 | 北京小米移动软件有限公司 | Method for detecting human face and device |
| CN107563347A (en) * | 2017-09-20 | 2018-01-09 | 南京行者易智能交通科技有限公司 | A kind of passenger flow counting method and apparatus based on TOF camera |
| CN108596128A (en) * | 2018-04-28 | 2018-09-28 | 京东方科技集团股份有限公司 | Object identifying method, device and storage medium |
| CN109615647A (en) * | 2018-10-24 | 2019-04-12 | 北京升哲科技有限公司 | Object detection method and device |
-
2018
- 2018-10-24 CN CN201811242524.6A patent/CN109584297A/en not_active Withdrawn
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102509074A (en) * | 2011-10-18 | 2012-06-20 | Tcl集团股份有限公司 | Target identification method and device |
| US20160364912A1 (en) * | 2015-06-15 | 2016-12-15 | Electronics And Telecommunications Research Institute | Augmented reality-based hand interaction apparatus and method using image information |
| CN106778614A (en) * | 2016-12-16 | 2017-05-31 | 中新智擎有限公司 | A kind of human body recognition method and device |
| CN107368810A (en) * | 2017-07-20 | 2017-11-21 | 北京小米移动软件有限公司 | Method for detecting human face and device |
| CN107563347A (en) * | 2017-09-20 | 2018-01-09 | 南京行者易智能交通科技有限公司 | A kind of passenger flow counting method and apparatus based on TOF camera |
| CN108596128A (en) * | 2018-04-28 | 2018-09-28 | 京东方科技集团股份有限公司 | Object identifying method, device and storage medium |
| CN109615647A (en) * | 2018-10-24 | 2019-04-12 | 北京升哲科技有限公司 | Object detection method and device |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109615647A (en) * | 2018-10-24 | 2019-04-12 | 北京升哲科技有限公司 | Object detection method and device |
| CN110207951A (en) * | 2019-05-23 | 2019-09-06 | 北京航空航天大学 | A kind of aircraft cable support assembled state detection method of view-based access control model |
| CN110207951B (en) * | 2019-05-23 | 2020-09-08 | 北京航空航天大学 | Vision-based aircraft cable bracket assembly state detection method |
| CN110220456A (en) * | 2019-06-26 | 2019-09-10 | 浙江大学 | A kind of hand-held box sizes measuring device and measuring method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10462362B2 (en) | Feature based high resolution motion estimation from low resolution images captured using an array source | |
| US8406510B2 (en) | Methods for evaluating distances in a scene and apparatus and machine readable medium using the same | |
| CN111127318B (en) | Panoramic image splicing method in airport environment | |
| CN108921781B (en) | Depth-based optical field splicing method | |
| US9048153B2 (en) | Three-dimensional image sensor | |
| CN107154014B (en) | Real-time color and depth panoramic image splicing method | |
| WO2018133641A1 (en) | A locating method and system | |
| CN108734657B (en) | Image splicing method with parallax processing capability | |
| CN105551050B (en) | A kind of image depth estimation method based on light field | |
| CN109584297A (en) | Object detection method and device | |
| CN108776980A (en) | A kind of scaling method towards lenticule light-field camera | |
| CN105869115B (en) | A kind of depth image super-resolution method based on kinect2.0 | |
| CN109446978B (en) | Method for tracking moving target of airplane based on staring satellite complex scene | |
| CN108156383B (en) | High-dynamic billion pixel video acquisition method and device based on camera array | |
| Zhang et al. | Synthetic aperture based on plenoptic camera for seeing through occlusions | |
| CN106471804A (en) | Method and device for picture catching and depth extraction simultaneously | |
| JP5900017B2 (en) | Depth estimation apparatus, reconstructed image generation apparatus, depth estimation method, reconstructed image generation method, and program | |
| CN112465702B (en) | Synchronous self-adaptive splicing display processing method for multi-channel ultrahigh-definition video | |
| CN108604371A (en) | Imaging system and method | |
| CN109615647A (en) | Object detection method and device | |
| KR101777696B1 (en) | Method For Subpixel Localization Of Edges Based on Areal Symmetry | |
| Pham et al. | Algorithm for military object detection using image data | |
| CN106683128B (en) | Sub-pixel registration method for airport runway image | |
| CN109961393A (en) | Subpixel registration and splicing based on interpolation and iteration optimization algorithms | |
| CN110956640B (en) | Heterogeneous image edge point detection and registration method |
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 | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20190405 |