CN109636835A - Foreground target detection method based on template light stream - Google Patents
Foreground target detection method based on template light stream Download PDFInfo
- Publication number
- CN109636835A CN109636835A CN201811536946.4A CN201811536946A CN109636835A CN 109636835 A CN109636835 A CN 109636835A CN 201811536946 A CN201811536946 A CN 201811536946A CN 109636835 A CN109636835 A CN 109636835A
- Authority
- CN
- China
- Prior art keywords
- template
- optical flow
- foreground target
- frame
- picture
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/269—Analysis of motion using gradient-based methods
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of foreground target detection methods based on template light stream, are related to computer vision target detection technique field, and technical solution is, including S1, selection original template;Optical flow field in S2, acquisition template picture and subsequent video stream between each frame picture;S3, the optical flow field that S2 is got is calculated, counts the vector length size of optical flow field, when statistical result is greater than threshold value, determine that present frame has the target different from template;S3, template renewal;At interval of the timeT, setting present frame picture is new template.The beneficial effects of the present invention are: optical flow method is lower to machine configuration requirement, save the cost.A large amount of tape label data are not needed, so the object detection method based on deep learning that compares, technically simple, are saved the time, it is practical.By more new template, it is adapted to the slowly varying situation of video background image.
Description
Technical field
The present invention relates to computer vision target detection technique field, in particular to a kind of prospect mesh based on template light stream
Mark detection method.
Background technique
Foreground target detection in quasi-static background belongs to computer vision scope, all has in all trades and professions huge
Practical application request, such as workshop, railroad track, airport hardstand, household safety-protection environment all swarm into given zone with detection
The target in domain is as core function.
Research field is detected in general target, in recent years fast-developing depth learning technology, met in performance real
Border use demand, these technical methods include but is not limited to YOLO (you only look once), SSD (Single Shot
MultiBox Detector).Object detection method based on deep learning generally needs a large amount of tape label data (to expend very much
Time and fund) model pre-training is carried out, while the high hardware environment (high performance video cards etc.) configured is needed, these difficulty limitations
It is in the application to cost sensitivity industry.
In view of this, how to realize that target detection becomes one and valuable studies a question efficiently at low cost.
Summary of the invention
In order to achieve the above-mentioned object of the invention, in view of the above technical problems, before the present invention provides a kind of light stream based on template
Scape object detection method.
Its technical solution is that this method is analyzed based on the optical flow field vector statistics of the every frame picture of video flowing, to detect prospect
Target, comprising:
S1, original template is chosen;
Within a bit of time of the initial phase of video flowing, a frame picture is randomly assigned as template, if the picture
In existing foreground target to be detected, then current template cancels, when needing to wait in video flowing there is no foreground target, again
Set template;
Optical flow field in S2, acquisition template picture and subsequent video stream between each frame picture;
S3, the optical flow field that S2 is got is calculated, counts the vector length size of optical flow field, when statistical result is greater than threshold value
When, determine that present frame has the target different from template;According to the experimental data given threshold of pre-training.
S4, template renewal;
When deviating or vibrating there are light variation, video camera, picture and template picture in live video stream
Difference will gradually amplify, and then may cause error detection, to solve this problem, need timely replacement template.Specifically, at interval of
Time T, if foreground target is not present in present frame, setting present frame picture is new template, if present frame there are foreground target,
It then postpones by video flowing, until more new template when foreground target is not present in present frame.
Preferably, in the S3, the statistical method of optical flow field are as follows:
Using classical optical flow method (to improve computational efficiency, preferentially using pyramid Lucas Kanade optical flow method), will regard
The template image that every frame image and S1 in frequency stream obtain carries out optical flow computation, calculates all light stream vectors in optical flow fieldLength
The average value of degreeThat is:
Wherein, n is the total quantity of light stream vector in optical flow field;I is light stream vector subscript, indicates corresponding i-th of light stream
Vector.
When light stream vector average valueWhen sharply increasing, determine that there are foreground targets.This is because the frame picture and template
There are great differences for picture, and lead to that optical flow method is calculated is optical flow field (some feature in template picture background of mispairing
The light stream of mispairing is constituted in point and live video stream between some incoherent characteristic point of foreground target), it cannot reflect light
The misalignment of stream method same substance point in general sense, can but be used to judge whether there is new object.
Preferably, a threshold value V is set by pre-training, whenWhen determine there are foreground targets.
Preferably, in the S4, when the present frame of video flowing and the time interval t of template are greater than preset value T, setting is current
The picture of frame is template.
Preferably, when the S4 more new template, detection present frame first whether there is foreground target;If result is
"No" is then new template using present frame picture;
If result be "Yes" if wouldn't more new template, continue to test next frame, until a certain frame in do not occur prospect mesh
When mark, use the frame picture as new template.
Preferably, template renewal time interval T is set according to the experimental data of pre-training.
Technical solution provided in an embodiment of the present invention has the benefit that optical flow method is lower to machine configuration requirement,
Save the cost.A large amount of tape label data are not needed, so the object detection method based on deep learning that compares, technology letter
It is single, it saves the time, it is practical.By more new template, it is adapted to the slowly varying situation of video background image.
Detailed description of the invention
Fig. 1 is the original template display diagram of the embodiment of the present invention.
Fig. 2 be the embodiment of the present invention without foreground target when optical flow field display diagram.
Fig. 3 is optical flow field display diagram when having a foreground target of the embodiment of the present invention.
Fig. 4 is the new template display diagram of the embodiment of the present invention.
Fig. 5 is the foreground target detection method flow chart based on template light stream of the embodiment of the present invention.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.Certainly, described herein specific examples are only used to explain the present invention, is not used to
Limit the present invention.
It should be noted that in the absence of conflict, the feature in embodiment and embodiment in the invention can
To be combined with each other.
In the description of the invention, it is to be understood that term " center ", " longitudinal direction ", " transverse direction ", "upper", "lower",
The orientation or positional relationship of the instructions such as "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outside" is
It is based on the orientation or positional relationship shown in the drawings, is merely for convenience of description the invention and simplifies description, rather than indicate
Or imply that signified device or element must have a particular orientation, be constructed and operated in a specific orientation, therefore cannot understand
For the limitation to the invention.In addition, term " first ", " second " etc. are used for description purposes only, and should not be understood as indicating
Or it implies relative importance or implicitly indicates the quantity of indicated technical characteristic." first ", " second " etc. are defined as a result,
Feature can explicitly or implicitly include one or more of the features.In the description of the invention, unless separately
It is described, the meaning of " plurality " is two or more.
In the description of the invention, it should be noted that unless otherwise clearly defined and limited, term " peace
Dress ", " connected ", " connection " shall be understood in a broad sense, for example, it may be being fixedly connected, may be a detachable connection, or integrally
Connection;It can be mechanical connection, be also possible to be electrically connected;Can be directly connected, can also indirectly connected through an intermediary,
It can be the connection inside two elements.For the ordinary skill in the art, on being understood by concrete condition
State concrete meaning of the term in the invention.
Embodiment 1
Referring to Fig. 1 to Fig. 5, the present invention provides a kind of foreground target detection method based on template light stream, to detect train
For whether there is train on track, detailed description of the present invention operating procedure.
Pre-training is carried out, real history video data is chosen, sets template, the light between calculation template and subsequent video frame
Flow vector, artificial judgment whether there is foreground target, and statistics exists respectively and there is no the light under foreground target both of these case
Flow field vector average value takes the average value of two end values as threshold value because the two results differ greatly.When detection
Between when continuing for some time t, template respectively with exist and there is no the matched light stream vector statistical values of the video frame of foreground target
Will be fairly close, so that there is mistake in matching result, to avoid this unfavorable as a result, must be when time interval is less than t, in time more
New template, specifically, settable template renewal interval T are approximately equal to the half of t.
Step 1 chooses original template.
After artificially judging that target to be detected is not present in current video, a frame is randomly selected as original template, Fig. 1
For template example:
Step 2, optical flow field is calculated, foreground target is detected.
Using pyramid Lucas Kanade optical flow method, the light stream in video flowing between every frame picture and template picture is calculated
?.Fig. 2 and Fig. 3 is respectively foreground target to be not present and the case where there are foreground targets, the statistics for calculating light stream vector length is flat
Mean value, as a result as shown in following two table.
(foreground target is not present) in 1 light stream vector length of table statistics
2 light stream vector length of table counts (there are foreground targets)
For the case where there are foreground targets, because the background mismatch in the foreground target and template of present frame causes to occur
A large amount of mistake light stream vector, with respect to no prospect target conditions, the length of these vectors is very big, need to only select suitable threshold value
V can be distinguished with the presence or absence of foreground target.For example V=1.0 is taken, and when being greater than this threshold value, that is, determine that there are foreground targets, it is on the contrary
Foreground target is then not present.
Step 3, more new template.
For the slowly varying situation of reply background picture, template renewal algorithm is introduced.At interval of the time, check that present frame is
No there are foreground targets, the use of present frame picture are new template if result is "No", wouldn't if result is "Yes"
More new template continues to test next frame, until using the frame picture as new mould when not occurring foreground target in a certain frame
Plate.Fig. 4 is updated template example (brightness is different from original template).
The foregoing is merely presently preferred embodiments of the present invention, is not intended to limit the invention, it is all in spirit of the invention and
Within principle, any modification, equivalent replacement, improvement and so on be should all be included in the protection scope of the present invention.
Claims (6)
1. the foreground target detection method based on template light stream, optical flow field vector statistics of this method based on the every frame picture of video flowing
Analysis, to detect foreground target characterized by comprising
S1, original template is chosen;
In the initial phase of video flowing, be randomly assigned a frame picture as template, if in the picture it is existing it is to be detected before
Scape target needs to wait for setting template again there is no when foreground target in video flowing then current template cancels;
Optical flow field in S2, acquisition template picture and subsequent video stream between each frame picture;
S3, the optical flow field that S2 is got is calculated, counts the vector length size of optical flow field, when statistical result is greater than threshold value, sentenced
There is the target different from template in settled previous frame;
S4, template renewal;At interval of time T, if foreground target is not present in present frame, setting present frame picture is new mould
Plate is postponed if there are foreground targets for present frame by video flowing, until more new template when foreground target is not present in present frame.
2. the foreground target detection method according to claim 1 based on template light stream, which is characterized in that in the S3,
The statistical method of optical flow field are as follows:
The template image of every frame image and S1 acquisition in video flowing is subjected to optical flow computation, calculates all light streams arrows in optical flow field
AmountLength average valueAccording to formula:
Wherein, n is the total quantity of light stream vector in optical flow field;
When light stream vector average valueWhen sharply increasing, determine that there are foreground targets.
3. the foreground target detection method according to claim 2 based on template light stream, which is characterized in that pass through pre-training
A threshold value V is set, whenWhen determine there are foreground targets.
4. the foreground target detection method according to claim 3 based on template light stream, which is characterized in that in the S4,
When the present frame of video flowing and the time interval t of template are greater than preset value T, the picture that present frame is arranged is template.
5. based on the foreground target detection method as claimed in claim 4 based on template light stream, which is characterized in that the S4 updates
When template, detection present frame first whether there is foreground target;It the use of present frame picture is new mould if result is "No"
Plate;If result is "Yes" wouldn't more new template, continue to test next frame, until in a certain frame when not occurring foreground target,
Use the frame picture as new template.
6. based on the foreground target detection method as claimed in claim 4 based on template light stream, which is characterized in that according to pre-training
Experimental data set template renewal time interval T.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811536946.4A CN109636835B (en) | 2018-12-14 | 2018-12-14 | Foreground object detection method based on template optical flow |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811536946.4A CN109636835B (en) | 2018-12-14 | 2018-12-14 | Foreground object detection method based on template optical flow |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109636835A true CN109636835A (en) | 2019-04-16 |
CN109636835B CN109636835B (en) | 2023-07-04 |
Family
ID=66074326
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811536946.4A Active CN109636835B (en) | 2018-12-14 | 2018-12-14 | Foreground object detection method based on template optical flow |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109636835B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110233967A (en) * | 2019-06-20 | 2019-09-13 | 漳州智觉智能科技有限公司 | Mould template image generation system and method |
CN110264458A (en) * | 2019-06-20 | 2019-09-20 | 漳州智觉智能科技有限公司 | Mold monitoring system and method |
CN110675369A (en) * | 2019-04-26 | 2020-01-10 | 深圳市豪视智能科技有限公司 | Coupling mismatch detection method and related equipment |
CN111754550A (en) * | 2020-06-12 | 2020-10-09 | 中国农业大学 | Method and device for detecting dynamic barrier in motion state of agricultural machine |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102800106A (en) * | 2012-06-29 | 2012-11-28 | 刘怡光 | Self-adaptation mean-shift target tracking method based on optical flow field estimation |
US20150139484A1 (en) * | 2013-11-19 | 2015-05-21 | Xerox Corporation | Time scale adaptive motion detection |
CN104966305A (en) * | 2015-06-12 | 2015-10-07 | 上海交通大学 | Foreground detection method based on motion vector division |
CN105654522A (en) * | 2015-12-30 | 2016-06-08 | 青岛海尔股份有限公司 | Frosting detection method and frosting detection system for evaporator |
CN105894530A (en) * | 2014-12-11 | 2016-08-24 | 深圳市阿图姆科技有限公司 | Detection and tracking solution scheme aiming at motion target in video |
CN106874949A (en) * | 2017-02-10 | 2017-06-20 | 华中科技大学 | A kind of moving platform moving target detecting method and system based on infrared image |
-
2018
- 2018-12-14 CN CN201811536946.4A patent/CN109636835B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102800106A (en) * | 2012-06-29 | 2012-11-28 | 刘怡光 | Self-adaptation mean-shift target tracking method based on optical flow field estimation |
US20150139484A1 (en) * | 2013-11-19 | 2015-05-21 | Xerox Corporation | Time scale adaptive motion detection |
CN105894530A (en) * | 2014-12-11 | 2016-08-24 | 深圳市阿图姆科技有限公司 | Detection and tracking solution scheme aiming at motion target in video |
CN104966305A (en) * | 2015-06-12 | 2015-10-07 | 上海交通大学 | Foreground detection method based on motion vector division |
CN105654522A (en) * | 2015-12-30 | 2016-06-08 | 青岛海尔股份有限公司 | Frosting detection method and frosting detection system for evaporator |
CN106874949A (en) * | 2017-02-10 | 2017-06-20 | 华中科技大学 | A kind of moving platform moving target detecting method and system based on infrared image |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110675369A (en) * | 2019-04-26 | 2020-01-10 | 深圳市豪视智能科技有限公司 | Coupling mismatch detection method and related equipment |
CN110675369B (en) * | 2019-04-26 | 2022-01-14 | 深圳市豪视智能科技有限公司 | Coupling mismatch detection method and related equipment |
CN110233967A (en) * | 2019-06-20 | 2019-09-13 | 漳州智觉智能科技有限公司 | Mould template image generation system and method |
CN110264458A (en) * | 2019-06-20 | 2019-09-20 | 漳州智觉智能科技有限公司 | Mold monitoring system and method |
CN110233967B (en) * | 2019-06-20 | 2021-10-01 | 漳州智觉智能科技有限公司 | Mold template image generation system and method |
CN110264458B (en) * | 2019-06-20 | 2023-01-06 | 漳州智觉智能科技有限公司 | Mold monitoring system and method |
CN111754550A (en) * | 2020-06-12 | 2020-10-09 | 中国农业大学 | Method and device for detecting dynamic barrier in motion state of agricultural machine |
CN111754550B (en) * | 2020-06-12 | 2023-08-11 | 中国农业大学 | Method and device for detecting dynamic obstacle in movement state of agricultural machine |
Also Published As
Publication number | Publication date |
---|---|
CN109636835B (en) | 2023-07-04 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109636835A (en) | Foreground target detection method based on template light stream | |
US10198623B2 (en) | Three-dimensional facial recognition method and system | |
CN105405154B (en) | Target object tracking based on color-structure feature | |
CN103208008B (en) | Based on the quick adaptive method of traffic video monitoring target detection of machine vision | |
CN111612820A (en) | Multi-target tracking method, and training method and device of feature extraction model | |
Lee et al. | Place recognition using straight lines for vision-based SLAM | |
EP4062374A1 (en) | Fish biomass, shape, size, or health determination | |
CN103942525A (en) | Real-time face optimal selection method based on video sequence | |
CN103530599A (en) | Method and system for distinguishing real face and picture face | |
WO2013053159A1 (en) | Method and device for tracking vehicle | |
CN110414376B (en) | Method for updating face recognition model, face recognition camera and server | |
CN103778409A (en) | Human face identification method based on human face characteristic data mining and device | |
CN102243765A (en) | Multi-camera-based multi-objective positioning tracking method and system | |
CN109145742A (en) | A kind of pedestrian recognition method and system | |
CN105701467A (en) | Many-people abnormal behavior identification method based on human body shape characteristic | |
CN110287907B (en) | Object detection method and device | |
CN110288627A (en) | One kind being based on deep learning and the associated online multi-object tracking method of data | |
CN106612457B (en) | Video sequence alignment schemes and system | |
CN103164693A (en) | Surveillance video pedestrian detection matching method | |
CN109711267A (en) | A kind of pedestrian identifies again, pedestrian movement's orbit generation method and device | |
CN111950345A (en) | Camera identification method and device, electronic equipment and storage medium | |
CN109684990A (en) | A kind of behavioral value method of making a phone call based on video | |
KR101214858B1 (en) | Moving object detecting apparatus and method using clustering | |
CN116681729B (en) | Anti-occlusion multi-target tracking method based on track prediction | |
CN101877135A (en) | Moving target detecting method based on background reconstruction |
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 |