CN108010057A - Target edit methods in a kind of interactive mode rail track real scene video - Google Patents
Target edit methods in a kind of interactive mode rail track real scene video Download PDFInfo
- Publication number
- CN108010057A CN108010057A CN201711223330.7A CN201711223330A CN108010057A CN 108010057 A CN108010057 A CN 108010057A CN 201711223330 A CN201711223330 A CN 201711223330A CN 108010057 A CN108010057 A CN 108010057A
- Authority
- CN
- China
- Prior art keywords
- target
- frame
- video
- selects
- edited
- 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
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
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
- G06F16/71—Indexing; Data structures therefor; Storage structures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
- G06V20/42—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items of sport video content
Abstract
Target edit methods in a kind of interactive mode rail track real scene video, belong to image and method for processing video frequency.This method is based on visualization interface, and after artificial frame choosing is completed in a small number of several two field pictures to some target, position of the target in other frames is tracked out using the method for maximizing local feature similarity.Position and quantity by adjusting the choosing of artificial frame, can obtain high-precision target following result.In the template image that the target is selected in the target location insertion that tracking obtains, the editor to target is completed.
Description
Technical field
The invention belongs to image and method for processing video frequency, and in particular to target in a kind of interactive mode rail track real scene video
Edit methods.
Background technology
The driving training of train operator is the key component of railroader's training, and ensures train driving safety
Core.Driving training should enable to driver to be familiar with different kinds of railways mark, skilled and various situations are setup flexibly.For up to
To this target, it is necessary to real driving environment is provided in training process, and video capture technology ripe at present can be with
Smooth and real train driving scene is obtained, training contents can be effectively enriched and simulate real driving experience.
The railway event included in the rail track real scene video that actually photographed is extremely limited, it is impossible to is met flexible
Various examination paper demand.Therefore video efficiently accurately edit, be the key problem in technology for meeting training requirement.In view of row
The running scene changes of car are sufficiently complex, and it is unstable that full automatic editor is carried out to railway target.
The content of the invention
The present invention proposes target edit methods in a kind of interactive rail track real scene video.It is intended to simplify rail track
All kinds of targets in real scene video.This method comprises the following steps：
（1）Video to be edited is decoded and played；
（2）Determine the time that target occurs：Watch video to be edited and determine the time range that target to be edited occurs；
（3）Frame selects target：In the time range that a certain target to be edited occurs, in some two field pictures to its position into pedestrian
Work frame selects, these have selected the two field picture of target to be known as frame and have selected frame by frame.Same target corresponds to one group of frame and selects frame, these frames select frame to exist
It is not adjacent in sequential, but is spaced two field pictures tens of or even up to a hundred.Frame select the number of target according to target zone and
Tracking quality and determine.For the engine video frequency of normal speed traveling, the frame that most targets need selects number on ten left sides
It is right；
（4）Automatic Target Following：Containing the video frame that substantial amounts of target location is unknown between frame selects frame, it is known as unknown frame.Need
The target location in frame is selected to track out the target location in unknown frame by frame.The track algorithm used can be it is arbitrary,
Such as the position of target in unknown frame, common feature such as color characteristic, shape can be obtained by maximizing the similarity of feature
Feature, textural characteristics, partial descriptions subcharacter, similarity measurement such as Euclidean distance, mahalanobis distance, Histogram distance, phase relation
Number.Enter next step if meeting quality requirement, otherwise return to step（3）；
（5）Selected template image：The edit effect performed as needed to some target, selects some preprepared template
Image, the type such as RGBA images of template image；
（6）Target editor：Contain the image of target area for each frame, by the size scaling of template image to and target area
It is identical, it is inserted into target area, is not processed for the two field picture for not containing target area；
（7）Video after editor：Original two field picture, the video after being edited are replaced using the two field picture after target editor.
The present invention utilizes the method for tracking target of image processing field, on the basis of a small amount of artificial frame choosing of target intactly
The target is tracked out in the appearance position of each frame of video.With reference to the modification demand of the target, you can be inserted into corresponding position
The template image of setting, completes the editor to railway target.This method can obtain standard on the basis of a small amount of artificial frame choosing
True tracking result, and less calculation amount ensure that higher operational efficiency.
The beneficial effects of the invention are as follows：A small amount of mark need to be only carried out to target, avoids and changes the numerous and diverse of video frame by frame
Work, and abundant driving scene can be provided for the training of train driving.
Brief description of the drawings
In order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, below will be to embodiment or existing
There is attached drawing needed in technology description to be briefly described, it should be apparent that, drawings in the following description are this hairs
Some bright embodiments, for those of ordinary skill in the art, without having to pay creative labor, can be with
Other attached drawings are obtained according to these attached drawings.
Fig. 1 is the two field picture containing semaphore in video to be edited；
Fig. 2 is to be directed to Fig. 1, carries out the rectangle frame that artificial frame selects to semaphore under magnification（Blueness）；
Fig. 3 is display result of the rectangle frame that frame selects in entire image after the frame of completion Fig. 2 selects process；
Fig. 4 is after the frame choosing of Fig. 3 is completed, and the rectangle frame result that frame next time selects is carried out after being spaced some frames（Rectangle
Frame）；
Fig. 5 is after Automatic Target Following is completed, and the corresponding tracking result of a frame between frame is selected in the artificial frame of Fig. 3 and Fig. 4
（Green rectangle frame）；
Fig. 6 is obtained edited result after greenish-yellow lamp template image is inserted into target location in the corresponding two field pictures of Fig. 5.It is special
It is RGBA images not point out greenish-yellow lamp template image, wherein top lamp is green, intermediate lamp is grey, and lower section lamp is yellow, on
It is not 0 to state the pixel A channel value in the region of three lamps, and rest of pixels A channel value is 0；
Fig. 7 flow charts of the present invention.
Embodiment
The invention will be further described below.
Target edit methods in a kind of interactive mode rail track real scene video, include the following steps：
（1）Video to be edited is decoded and played：Video to be edited is decoded, and obtained two field picture will be decoded by video to be edited
Frame per second shown, reach play video effect, give tacit consent to from 0 moment and commence play out.User can select any instant work
At the time of currently to commence play out, the directly moment corresponding frame starts to decode after completing setting；
（2）Determine the time that target occurs：The video to be edited played is watched, identifies the target of required editor, these targets
It can continue the presence of a period of time in video；Fig. 1 is the two field picture in the range of some semaphore target occurs.Next should
Illustrated exemplified by target；
（3）Frame selects target：In the time range that current goal occurs, suspend the broadcasting of video at any one time, in pause shape
Current frame image is extracted under state.The two field picture extracted is zoomed in and out support user and translation, accurately to be looked into
See.Mesh target area is selected in the two field picture upper ledge extracted（Draw rectangle frame）, region that the frame is selected is regarded as the frame figure
The boundary rectangle of required modification target as in.
It is assumed that determine then be directed to the two field picture of Fig. 1 to the semaphore into edlin to the semaphore target in Fig. 1
Position carries out a frame choosing, and the mode of frame choosing is to draw rectangle frame using mouse.Fig. 2 is illustrated in the state of two field picture is amplified
The result of rectangle circle choosing is carried out to semaphore（Rectangle frame）.Fig. 3 is illustrated be switched to original broadcast state after, in view picture frame figure
As the rectangle circle of upper displaying selects result（Rectangle frame）.
If in order to which the video segment occurred to the semaphore carries out completely modification, it is necessary to be carried out on different two field pictures
Dry frame choosing.Fig. 4 is illustrated after the frame for carrying out Fig. 1 to Fig. 3 selects process, the result that frame selects next time（Rectangle frame）.
It is that non-frame selects two field picture between Fig. 3 and the two field picture of Fig. 4, Fig. 5 corresponds to a wherein frame therein
Repeat the above steps, frame choosing can be carried out on different frames to the semaphore target.Two details are pointed out, when
Select how many frames to carry out frame choosing and frame choosing is carried out on which frame by user's unrestricted choice.Second, it is proposed that user
First select less frame to carry out the drafting of rectangle frame, then perform the（4）The Automatic Target Following of step, if tracking result is unsatisfactory for
The more frames of demand reselection carry out frame choosing, can so further improve the effect of target following；
（4）Automatic Target Following：Complete step（3）Afterwards, the position of current goal is only determined on some two field pictures selected by frame
Put, the position of the target on other two field pictures is obtained according to these known locations, which is target following；
Detailed process is：In the two field picture of each manually frame choosing, the feature description of image, the present invention carry in calculation block favored area
It will be described in the example of confession using the gray value in region as feature；For any untagged frame image, if the two field picture when
Between on not between two artificial frames select two field picture, then do not calculate the target location on the image.
Otherwise, it determines the previous manually frame of the frame in time selects two field picture and the artificial frame of the latter to select two field picture.So
Selected afterwards in the two artificial frames near the mark position in two field picture, series of rectangular image block is selected according to certain intervals.Often
The size of a image block randomly selects, and restrictive condition is greater than two artificial frames and selects smaller frame in frame to select size, and is less than
Two artificial frames select larger frame in frame to select size, while aspect ratio is equal to the aspect ratio that two artificial frames select frame center to select rectangle
Average.
Its score is calculated one by one for this series of images block, calculating process is related to the similarity measure of feature description, this
Calculation of the related coefficient as similarity will be used by inventing in the example provided.To any image block, scale it first
It is identical with previous manually frame constituency domain sizes for size, obtain after scaling the feature description of image and with previous manually frame favored area
The feature description of interior image calculates similarity；It is identical with latter manually frame constituency domain sizes that the image block is scaled size again,
Obtain the feature description of image after scaling and calculate similarity with the feature description of image in latter artificial frame favored area.Gained two
The average of a similarity is the score of the image block.
In this series of images block, the corresponding position of image block of highest scoring, i.e., as mesh in the untagged frame image
Position where marking.
Above-mentioned tracking process is further illustrated with reference to attached drawing.Green rectangle frame in Fig. 5 be perform Automatic Target Following after,
In the target rectangle frame that the frame obtains.What the example used in object tracking process is retouched using the gray value in region as feature
State, in a manner of related coefficient is calculates similarity, the highest image block of similarity, its position are selected in the image block of candidate
The target location of two field picture is selected as the non-frame.Particularly point out, grey value characteristics description used herein and related coefficient phase
It is processing strategy basic in target following like degree, is described using other classical features（Such as color histogram feature, angle point
Feature, partial descriptions subcharacter etc.）, or other classical similarities（Such as the Euclidean distance, mahalanobis distance, angle of feature vector
Cosine value etc.）, can obtain similar tracking result.
After the non-frame of each width occurred to signal lamp target selects two field picture to complete target following, you can obtain signal lamp mesh
The position being marked on each frame.
（5）Selected template image：Template image needs to complete in advance, and template image can be RGB triple channel images,
Can also be RGBA four-way images（Wherein A channel determines the transparency of image）, above two image is general image
Type.For a target, select a template image and be used for follow-up edit operation；
（6）Target editor：Selected template image is inserted into the region of current goal in each frame, mould is scaled in insertion process
Plate image makes it be consistent with target area.As described above, the mesh target area position that either frame selects in each frame, or with
Position obtained by track.In the case of template image is different from target area aspect ratio, system is to provide the plan of two kinds of auto zooms
Slightly：The aspect ratio for changing template image is consistent with target area；Or on the premise of keeping the aspect ratio of template image, make length/width
Equal to the length/width of target area, while width/length is less than the width/length of target area, and the template image after scaling is placed centrally
In target area.
Particularly point out the insertion details of two kinds of image types.For RGB triple channel images, insertion process is by target area
Interior image directly replaces with template image.For RGBA four-way images, the value according to each pixel A passage makes choice
Property processing, when A channel value is not 0, the value of the pixel in the target area is replaced with to the value of the pixel in template image；
Otherwise, the value of the pixel in target area is not changed.
By taking the semaphore target in attached drawing as an example, Fig. 6 is the signal location that greenish-yellow lamp template image is inserted into the frame
The image of gained, the image will be used for corresponding original two field picture in more new video.Every two field picture corresponding to the signal lamp is all
The insertion process is carried out, that is, completes the editor of the signal lamp in video（7）Video after editor：Step（3）Arrive（6）Describe
The editing process of either objective, repeats the above process the editor that can be completed to each target, and the video obtained after mark is
Complete the video of target editor.
In video used in the present invention, include but not limited to the high definition that high-definition camera is shot during train driving
Color video.Feature description during tracking, can be, but not limited to gray value, color histogram feature, corner feature,
Partial descriptions subcharacter.The similarity measure, can be, but not limited to the related coefficient of feature vector, Yi Jili during tracking
The methods of being calculated similarity with Pasteur's distance on rgb space, calculated similarity using the Euclidean distance in HIS space calculates
The similarity of region unit.The object edited, includes but not limited to various types of signal machine, marker, foreign matter in the railway system.
The present invention need to only carry out target a small amount of mark, avoid the miscellaneous work for changing video frame by frame, and can be
The training of train driving provides abundant driving scene.
Claims (1)
1. target edit methods in a kind of interactive mode rail track real scene video, it is characterised in that include the following steps：
（1）Video to be edited is decoded and played；
（2）Determine the time that target occurs：Watch video to be edited and determine the time range that target to be edited occurs；
（3）Frame selects target：In the time range that a certain target to be edited occurs, in some two field pictures to its position into pedestrian
Work frame selects, these have selected the two field picture of target to be known as frame and have selected frame by frame；Same target corresponds to one group of frame and selects frame, these frames select frame to exist
It is not adjacent in sequential, but is spaced two field pictures tens of or even up to a hundred；Frame select the number of target according to target zone and
Tracking quality and determine；For the engine video frequency of normal speed traveling, the frame that most targets need selects number on ten left sides
It is right；
（4）Automatic Target Following：Containing the video frame that substantial amounts of target location is unknown between frame selects frame, it is known as unknown frame；Need
The target location in frame is selected to track out the target location in unknown frame by frame；The track algorithm used can be it is arbitrary,
Such as the position of target in unknown frame, common feature such as color characteristic, shape can be obtained by maximizing the similarity of feature
Feature, textural characteristics, partial descriptions subcharacter, similarity measurement such as Euclidean distance, mahalanobis distance, Histogram distance, phase relation
Number；Enter next step if meeting quality requirement, otherwise return to step（3）；
（5）Selected template image：The edit effect performed as needed to some target, selects some preprepared template
Image, the type such as RGBA images of template image；
（6）Target editor：Contain the image of target area for each frame, by the size scaling of template image to and target area
It is identical, it is inserted into target area, is not processed for the two field picture for not containing target area；
（7）Video after editor：Original two field picture, the video after being edited are replaced using the two field picture after target editor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711223330.7A CN108010057A (en) | 2017-11-29 | 2017-11-29 | Target edit methods in a kind of interactive mode rail track real scene video |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711223330.7A CN108010057A (en) | 2017-11-29 | 2017-11-29 | Target edit methods in a kind of interactive mode rail track real scene video |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108010057A true CN108010057A (en) | 2018-05-08 |
Family
ID=62054461
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711223330.7A Pending CN108010057A (en) | 2017-11-29 | 2017-11-29 | Target edit methods in a kind of interactive mode rail track real scene video |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108010057A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109493367A (en) * | 2018-10-29 | 2019-03-19 | 浙江大华技术股份有限公司 | The method and apparatus that a kind of pair of target object is tracked |
| CN109635777A (en) * | 2018-12-24 | 2019-04-16 | 广东理致技术有限公司 | A kind of video data editing recognition methods and device |
| CN109640145A (en) * | 2018-12-24 | 2019-04-16 | 郑州畅想高科股份有限公司 | A kind of driving drilling method and device for track train |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103702040A (en) * | 2013-12-31 | 2014-04-02 | 广州华多网络科技有限公司 | Real-time video graphic decoration superposing processing method and system |
| CN104394313A (en) * | 2014-10-27 | 2015-03-04 | 成都理想境界科技有限公司 | Special effect video generating method and device |
| CN105635807A (en) * | 2015-12-30 | 2016-06-01 | 北京奇艺世纪科技有限公司 | Video editing method and apparatus |
| CN105959724A (en) * | 2016-05-24 | 2016-09-21 | 腾讯科技（深圳）有限公司 | Video data processing method and device |
-
2017
- 2017-11-29 CN CN201711223330.7A patent/CN108010057A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103702040A (en) * | 2013-12-31 | 2014-04-02 | 广州华多网络科技有限公司 | Real-time video graphic decoration superposing processing method and system |
| CN104394313A (en) * | 2014-10-27 | 2015-03-04 | 成都理想境界科技有限公司 | Special effect video generating method and device |
| CN105635807A (en) * | 2015-12-30 | 2016-06-01 | 北京奇艺世纪科技有限公司 | Video editing method and apparatus |
| CN105959724A (en) * | 2016-05-24 | 2016-09-21 | 腾讯科技（深圳）有限公司 | Video data processing method and device |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109493367A (en) * | 2018-10-29 | 2019-03-19 | 浙江大华技术股份有限公司 | The method and apparatus that a kind of pair of target object is tracked |
| CN109493367B (en) * | 2018-10-29 | 2020-10-30 | 浙江大华技术股份有限公司 | Method and equipment for tracking target object |
| CN109635777A (en) * | 2018-12-24 | 2019-04-16 | 广东理致技术有限公司 | A kind of video data editing recognition methods and device |
| CN109640145A (en) * | 2018-12-24 | 2019-04-16 | 郑州畅想高科股份有限公司 | A kind of driving drilling method and device for track train |
| CN109640145B (en) * | 2018-12-24 | 2021-08-06 | 郑州畅想高科股份有限公司 | Driving practicing method and device for rail train |
| CN109635777B (en) * | 2018-12-24 | 2022-09-13 | 广东理致技术有限公司 | Video data editing and identifying method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP2380358B1 (en) | Creation of depth maps from images | |
| CN102859675B (en) | Semiconductor fault analysis device and fault analysis method | |
| CN108010057A (en) | Target edit methods in a kind of interactive mode rail track real scene video | |
| CN101501776B (en) | Video object cut and paste | |
| CN109151439A (en) | A kind of the automatic tracing camera system and method for view-based access control model | |
| JP4658266B2 (en) | Horse position information analysis and display method | |
| CN108985229A (en) | A kind of intelligent advertisement replacement method and system based on deep neural network | |
| GB2416949A (en) | Insertion of additional content into video | |
| JP2009505553A (en) | System and method for managing the insertion of visual effects into a video stream | |
| CN102202224A (en) | Caption flutter-free method and apparatus used for plane video stereo transition | |
| CN107577981A (en) | A kind of road traffic index identification method and device | |
| CN104952256A (en) | Video information based method for detecting vehicles at intersection | |
| JP4156084B2 (en) | Moving object tracking device | |
| CN110420445B (en) | Wall ball training method and device based on augmented reality | |
| CN103489154A (en) | Suture line real-time regulating method based on target block detection | |
| CN109640145B (en) | Driving practicing method and device for rail train | |
| CN105100773B (en) | Three-dimensional video-frequency preparation method, three-dimensional view preparation method and manufacturing system | |
| CN107948570A (en) | A kind of rail track real scene video plays synchronous locomotive driving real time information source method | |
| US9807339B2 (en) | Frame rate conversion system | |
| CN113902765B (en) | Automatic semiconductor partitioning method based on panoramic segmentation | |
| CN109978916A (en) | Vibe moving target detecting method based on gray level image characteristic matching | |
| CN106231278A (en) | Method for processing video frequency and television system | |
| JPH04235589A (en) | Display of contents of animated picture | |
| CN104333726B (en) | A kind of interlace-removing method and system for transcoded video source | |
| CN112700485B (en) | Image depth information extraction 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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180508 |