CN104580825A - Method for identifying and tracking object in video - Google Patents
Method for identifying and tracking object in video Download PDFInfo
- Publication number
- CN104580825A CN104580825A CN201510041443.XA CN201510041443A CN104580825A CN 104580825 A CN104580825 A CN 104580825A CN 201510041443 A CN201510041443 A CN 201510041443A CN 104580825 A CN104580825 A CN 104580825A
- Authority
- CN
- China
- Prior art keywords
- frame
- block structure
- video
- block
- image
- 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
Abstract
The invention provides a method for identifying and tracking an object in a video. The method for identifying the object in the video includes the steps that a first frame of video bit stream is decoded, and a first frame of image and a block structure of the image are obtained, wherein the first frame is any frame in video bit streams, the block structure is composed of a plurality of blocks divided according to textural features of the image; the object in the first frame of image is identified; the object is matched with the block structure of the first frame, wherein the block structure corresponding to the object serves as an object structure related to the object; entropy decoding is carried out on followed frames of video bit streams, and block structures of the followed frames are obtained, wherein the followed frames are multiple continuous frames behind the first frame; whether a first block structure the same as the object structure exists in the followed frames of block structures or not is judged, and the first block structure represents the object if the first block structure exists in the followed frames of block structures. By means of the method, the calculation amount can be reduced, and the tracking instantaneity of the object in the video is improved.
Description
Technical field
The present invention relates to field of video image processing, in video the method for identification and tracing object.
Background technology
Video object tracking has that precision is high, good concealment and the strong advantage of intuitive.These advantages make Video object tracking be widely used in fields such as target range photoelectric measurement, astronomical observation equipment, weapon control system, laser communication system, traffic monitoring, scene analysis, population analysis, people counting, Gait Recognition, action recognition.According to the difference of tracked object information service condition, Vision Tracking can be divided into based on the Object tracking of Analysis of Contrast, based on the Object tracking of coupling and the Object tracking based on motion detection.Track algorithm based on Analysis of Contrast mainly utilizes the contrast difference of object and background, realizes the detection and tracking of object.Realize the location of object mainly through the characteristic matching between the frame of front and back based on the tracking of mating.Tracking based on motion detection mainly realizes the detection and tracking of object according to the difference between object motion and background motion.
For the video tracking algorithm based on coupling, in current video, in Object identifying or video, the technology of Object tracking is: first the video received is decoded, then treatment and analysis is carried out to the pixel of each two field picture in video, identify the object in video or the further object followed the tracks of in video.And such technology relates generally to video decode, and image procossing and graphical analysis scheduling algorithm are carried out to each two field picture, the computing capability of at substantial, be generally difficult to reach high frame per second and detect in real time, particularly along with the increase of video pixel, high performance CPU or GPU is more needed to go process.
Summary of the invention
The defect that the present invention exists to overcome above-mentioned prior art, provide a kind of and identify the method for object and the method for tracing object in video, it can reduce amount of calculation and improve object tracking real-time in video.
The invention provides a kind of method identifying object in video, comprise: the video code flow of the first frame is decoded, obtain image and the block structure thereof of described first frame, described first frame is the arbitrary frame in described video code flow, described block structure divided by the textural characteristics according to image multiple pieces form; Identify the object in the image of described first frame; The block structure of described object with described first frame is mated, will the block structure of described object be corresponded to as the object structure with described object association; Only carry out entropy decoding to the video code flow of subsequent frame, obtain the block structure of described subsequent frame, described subsequent frame is the continuous multiple frames after described first frame; And whether have first block structure identical with described object structure in the block structure judging described subsequent frame, if having, then described first block structure is represented described object.
Preferably, described block structure is the block structure of maximum coding unit or the block structure of macro block.
Preferably, described block structure divides according to the entropy of each piece in image and threshold value.
Preferably, image and the block structure thereof of each frame of video form described video code flow through entropy code, wherein, comprise the decoding of described video code flow: utilize the entropy corresponding with described entropy code to decode and to decode to described video code flow.
Preferably, described first frame is key frame, wherein, identifies that the object in the image of described first frame comprises: identify the object outline in the image of described first frame.
Preferably, described object is mated with the block structure of described first frame comprise: the block structure of described object outline with described first frame is mated.
Preferably, the edge of described object structure is contained in described object outline, and described object structure is block structure maximum in described object outline.
Preferably, the edge of described object structure comprises described object outline, and described object structure is the minimum block structure covering described object.
Preferably, the edge of described object structure is crossing with described object outline, and the size that block crossing with described object outline in described object structure covers the part of described object is at least more than or equal to 1/3rd of the size of described piece.
According to another aspect of the invention, a kind of method of tracing object in video is also provided, comprises: the object in above-mentioned method identification video; According to the object identified, follow the tracks of position and the motion of described object.
Compared with prior art, the present invention devises and utilizes the partial information in video code flow to realize identifying the object in frame of video and following the tracks of, it only needs to decode to key frame or the frame chosen arbitrarily, image procossing and analysis, and do not need to decode to each frame of whole video, image procossing and graphical analysis, greatly reduce calculating of the identification of object in video and tracking image process and analysis, and improve the real-time of Object identifying and tracking in video.Adopt the method for this invention to carry out the object in video following the tracks of the load that can reduce CPU or GPU, improve the speed of Object tracking in video, reduce energy consumption.
Accompanying drawing explanation
Describe its example embodiment in detail by referring to accompanying drawing, above-mentioned and further feature of the present invention and advantage will become more obvious.
Fig. 1 shows the flow chart identifying the method for object in video according to the embodiment of the present invention.
Fig. 2 shows the block structured schematic diagram according to the embodiment of the present invention.
Fig. 3 shows the schematic diagram of object according to a first embodiment of the present invention and block structure coupling.
Fig. 4 shows the schematic diagram of object according to a second embodiment of the present invention and block structure coupling.
Fig. 5 shows the schematic diagram of object according to a third embodiment of the present invention and block structure coupling.
Fig. 6 shows the schematic diagram of Object identifying according to a third embodiment of the present invention.
Embodiment
More fully example embodiment is described referring now to accompanying drawing.But example embodiment can be implemented in a variety of forms, and should not be understood to be limited to execution mode set forth herein; On the contrary, these execution modes are provided to make the present invention comprehensively with complete, and the design of example embodiment will be conveyed to those skilled in the art all sidedly.Reference numeral identical in the drawings represents same or similar structure, thus will omit the repeated description to them.
Fig. 1 shows the flow chart identifying the method for object in video according to the embodiment of the present invention.And particularly illustrate following steps:
Step S101, decodes to the video code flow of the first frame, obtains image and the block structure thereof of the first frame.Wherein, the first frame is the arbitrary frame in video code flow.Preferably, the first frame is the key frame of video code flow.Video code flow is preferably the video code flow of HEVC form, but not as limit.Multiple pieces of the different sizes that block structure is divided by the textural characteristics according to image form.As shown in Figure 2, block structure is made up of the block (namely coding unit Coding Unit) of multiple different size the block structure of a specific embodiment.The block generally divided than more rich place in minutia or textural characteristics is little, the block that place that is smooth or background divides is large, thus the block structure secondary indication morphosis of certain object and object, can use as the feature structure of object and object and carry out characteristic matching.
In a preference, block structure is the block structure of maximum coding unit.The image of each frame of video and block structure thereof form HEVC video code flow through entropy code.Such as, can in the following way image be divided:
First image is divided into multiple maximum coding unit (LCU), maximum coding unit can be the coding unit of 64*64 pixel.Remove the background noise of each maximum coding unit.For each maximum coding unit, calculate the entropy of the coding unit (CU) likely marked off in this maximum coding unit.Such as, for the maximum coding unit of 64*64 pixel, it can be divided into the coding unit (CU8*8) of the coding unit (CU64*64) of 1 64*64 pixel, the coding unit (CU32*32) of 4 32*32 pixels, the coding unit (CU16*16) of 16 16*16 pixels, 64 8*8 pixels.In the present embodiment, the entropy of calculating 85 coding units is needed.
According to the entropy of calculated coding unit and threshold value, the division rule of picture block structure determines whether described coding unit should be divided by continuation.If the entropy of one of them coding unit is less than first threshold, then this coding unit does not need to divide again; If the entropy of this coding unit is greater than Second Threshold, then this coding unit needs to continue to be divided.Or if the entropy of this coding unit is close to the mean value of the entropy of the coding unit likely marked off, then this coding unit can not divide again.In a preferred embodiment, if the entropy of a coding unit is less than 1.2 (first thresholds), then think that the entropy of this coding unit is very little and do not need to divide again, if the entropy of a coding unit is greater than 3.5 (Second Thresholds), then thinks that the entropy of this coding unit is very large and need to continue to be divided.Or, if the entropy of a coding unit the entropy of the coding unit likely marked off mean value ± scope of 0.15 within, then this coding unit entropy close to the entropy of the described coding unit likely marked off mean value and can not divide again.
Those skilled in the art can also use other partitioning algorithms, and image is divided into multiple pieces, do not repeat them here.
In a change case, block structure can also be the block structure of macro block, and preferably, the size of each macro block is 16*16 pixel, and it can be made up of the block of the block of the block of 16*16 pixel, 8*8 pixel, 4*4 pixel or their combination.Divide similar with the block structured of maximum coding unit, the block structure of macro block also divides according to the textural characteristics of image, thus the block structure secondary indication morphosis of certain object and object, can use as the feature structure of object and object and carry out characteristic matching.
Step S102, identifies the object in the image of the first frame.In one embodiment, the object in Target Recognition Algorithms recognition image can be utilized.Those skilled in the art can also use the object in other recognizer recognition images.Particularly, the multiple objects in Target Recognition Algorithms recognition image can be utilized, then select an object as object to be identified as required, or also can using multiple object as object to be identified.
Preferably, the object outline in the image of all kinds of algorithm identified first frame can be utilized.Such as, boundary tracking algorithm identified object outline is utilized.
Step S103, mates the block structure of the object identified with the first frame, will correspond to the block structure of object as the object structure with object association.
In a preference, the block structure of object outline with the first frame is mated, and determines object structure in the following way.In one embodiment, the edge of object structure is contained in described object outline, and object structure is block structure maximum in object outline.As shown in Figure 3, based on identifying object 200 and object outline 210 in previous step, block structure 100 mates with object 200, using be positioned at object outline 210 by the multiple pieces of block structures formed as object structure 300, wherein, the edge 310 of object structure 300 is positioned at object outline 210, and object structure 300 is block structures maximum in object outline 210.
In another embodiment, object outline 210 is positioned at the edge 310 ' of object structure 300 ', and object structure 300 ' is the minimum block structure covering object 200.As shown in Figure 4, based on identifying object 200 and object outline 210 in previous step, block structure 100 mates with object 200, using comprise object 200 by the multiple pieces of block structures formed as object structure 310 ', wherein, object structure 310 ' is the minimum block structure covering object 200.
In other embodiment, object structure 300 " edge 310 " crossing with object outline 210, object structure 300 " in the block crossing with object outline 210 size that covers the part of object 200 be at least more than or equal to 1/3rd of the size of this block.As shown in Figure 5, based on identifying object 200 and object outline 210 in previous step, block structure 100 mates with object 200.In order to make object structure 300 " edge 310 " press close to object outline 210 as far as possible, object structure 300 " and in the block crossing with object outline 210 size that covers the part of object 200 be more than or equal to 1/3rd of the size of this block.In a change case, object structure 300 can be made " in the block crossing with object outline 210 size that covers the part of object 200 be more than or equal to 1/2nd of the size of this block.
Those skilled in the art can also pass through other algorithm determination object structure, do not repeat them here.
Step S104, decodes to the video code flow of subsequent frame, and obtain the block structure of subsequent frame, subsequent frame is the continuous multiple frames after the first frame.In the present embodiment, the first frame is key frame, and subsequent frame is the continuous multiple frames after key frame.
In a preference, video code flow is form HEVC video code flow through entropy code.This comprises video code flow decoding: utilize the entropy decoding corresponding with entropy code to decode to video code flow.Such as, call entropy decoder module (CABAC or CAVLC) and only entropy decoding is carried out to subsequent frame, do not need each frame to be reconstructed into an image, only need to decode block structure (block structure 100 ' as shown in Figure 6).
Step S105, judges whether there be first block structure identical with object structure in the block structure of subsequent frame, if having, then by the first block structure representative object.
See Fig. 5 and Fig. 6, Fig. 6 illustrates the block structure 100 ' of subsequent frame, object structure 300 by determining in Fig. 5 " mate in block structure 100 ', the first block structure 400 in the block structure 100 ' of subsequent frame " with object structure 300 " identical, then the first block structure 400 " can representative object 200.
The present invention also provides the method for tracing object in video, after utilizing above-mentioned steps S101 to S105 to identify object, and the position of tracing object and motion.Such as, according to said method, in Fig. 5, object 200 is on the right side of image, and in figure 6, the block structure 400 of representative object 200 " on the left of image, and then known object 200 moves to the left from right side in key frame and subsequent frame thereof.
Preferably, the method for tracing object in video provided by the invention processes according to step S101 to S105 key frames all in video code flow and subsequent frame thereof, so in whole video the movement tendency and/or movement locus etc. of tracing object.In some change case, the method for tracing object in video provided by the invention only processes according to step S101 to S105 the Partial key frame in video code flow and subsequent frame (can comprise key frame) thereof.
Compared with prior art, the present invention devises and utilizes the partial information in video code flow to realize identifying the object in frame of video and following the tracks of, do not need to decode to each frame of whole video, image procossing and graphical analysis, greatly reduce calculating of the identification of object in video and tracking image process and analysis, and improve the real-time of Object identifying and tracking in video.Carry out following the tracks of the load that can reduce CPU or GPU to the object in video by the method for this invention, improve the speed of Object tracking in video, reduce energy consumption.
Below illustrative embodiments of the present invention is illustrate and described particularly.Should be appreciated that, the invention is not restricted to disclosed execution mode, on the contrary, the invention is intended to contain the various amendment and equivalent replacement that comprise within the scope of the appended claims.
Claims (10)
1. identify a method for object in video, it is characterized in that, comprising:
Decode to the video code flow of the first frame, obtain image and the block structure thereof of described first frame, described first frame is the arbitrary frame in described video code flow, and described block structure multiple pieces of being divided by the textural characteristics according to image form;
Identify the object in the image of described first frame;
The block structure of described object with described first frame is mated, will the block structure of described object be corresponded to as the object structure with described object association;
Carry out entropy decoding to the video code flow of subsequent frame, obtain the block structure of described subsequent frame, described subsequent frame is the continuous multiple frames after described first frame; And
Judge in the block structure of described subsequent frame, whether there be first block structure identical with described object structure, if having, then described first block structure is represented described object.
2. video frequency identifying method as claimed in claim 1, it is characterized in that, described block structure is the block structure of maximum coding unit or the block structure of macro block.
3. video frequency identifying method as claimed in claim 2, is characterized in that, described block structure divides according to the entropy of each piece in image and threshold value.
4. video frequency identifying method as claimed in claim 1, it is characterized in that, image and the block structure thereof of each frame of video form described video code flow through entropy code, wherein, comprise the decoding of described video code flow: utilize the entropy corresponding with described entropy code to decode and to decode to described video code flow.
5. video frequency identifying method as claimed in claim 1, it is characterized in that, described first frame is key frame, wherein, identifies that the object in the image of described first frame comprises: identify the object outline in the image of described first frame.
6. video frequency identifying method as claimed in claim 5, is characterized in that, mated by described object and comprise: mated by the block structure of described object outline with described first frame with the block structure of described first frame.
7. video frequency identifying method as claimed in claim 6, it is characterized in that, the edge of described object structure is contained in described object outline, and described object structure is block structure maximum in described object outline.
8. video frequency identifying method as claimed in claim 6, it is characterized in that, the edge of described object structure comprises described object outline, and described object structure is the minimum block structure covering described object.
9. video frequency identifying method as claimed in claim 6, it is characterized in that, the edge of described object structure is crossing with described object outline, and the size that block crossing with described object outline in described object structure covers the part of described object is at least more than or equal to 1/3rd of the size of described piece.
10. a method for tracing object in video, is characterized in that, comprising:
Utilize the object in the method identification video described in any one of claim 1 to 9;
According to the object identified, follow the tracks of position and the motion of described object.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510041443.XA CN104580825B (en) | 2015-01-28 | 2015-01-28 | The method for identifying in video and tracking object |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510041443.XA CN104580825B (en) | 2015-01-28 | 2015-01-28 | The method for identifying in video and tracking object |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104580825A true CN104580825A (en) | 2015-04-29 |
| CN104580825B CN104580825B (en) | 2018-05-08 |
Family
ID=53095935
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510041443.XA Active CN104580825B (en) | 2015-01-28 | 2015-01-28 | The method for identifying in video and tracking object |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104580825B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107491172A (en) * | 2017-08-16 | 2017-12-19 | 歌尔科技有限公司 | Body-sensing data capture method, device and electronic equipment |
| CN108881908A (en) * | 2018-05-28 | 2018-11-23 | 北方工业大学 | Quick partitioning based on coding unit Texture complication in Video coding |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7606435B1 (en) * | 2002-02-21 | 2009-10-20 | At&T Intellectual Property Ii, L.P. | System and method for encoding and decoding using texture replacement |
| CN102324030A (en) * | 2011-09-09 | 2012-01-18 | 广州灵视信息科技有限公司 | Target tracking method and system based on image block characteristics |
| CN103813169A (en) * | 2014-02-19 | 2014-05-21 | 北京大学 | Extensible object notation method and device for use in video coder/decoder |
-
2015
- 2015-01-28 CN CN201510041443.XA patent/CN104580825B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7606435B1 (en) * | 2002-02-21 | 2009-10-20 | At&T Intellectual Property Ii, L.P. | System and method for encoding and decoding using texture replacement |
| CN102324030A (en) * | 2011-09-09 | 2012-01-18 | 广州灵视信息科技有限公司 | Target tracking method and system based on image block characteristics |
| CN103813169A (en) * | 2014-02-19 | 2014-05-21 | 北京大学 | Extensible object notation method and device for use in video coder/decoder |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107491172A (en) * | 2017-08-16 | 2017-12-19 | 歌尔科技有限公司 | Body-sensing data capture method, device and electronic equipment |
| CN108881908A (en) * | 2018-05-28 | 2018-11-23 | 北方工业大学 | Quick partitioning based on coding unit Texture complication in Video coding |
| CN108881908B (en) * | 2018-05-28 | 2022-07-01 | 北方工业大学 | Coding unit texture complexity based fast partitioning in video coding |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104580825B (en) | 2018-05-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106231214B (en) | High-speed CMOS sensor image approximate lossless compression method based on adjustable macro block | |
| CN111355956B (en) | Deep learning-based rate distortion optimization rapid decision system and method in HEVC intra-frame coding | |
| US10602186B2 (en) | Systems and methods for mask based processing of a block of a digital image | |
| Zhang et al. | Low complexity HEVC INTRA coding for high-quality mobile video communication | |
| US20200007872A1 (en) | Video decoding method, video decoder, video encoding method and video encoder | |
| US11727255B2 (en) | Systems and methods for edge assisted real-time object detection for mobile augmented reality | |
| Dey et al. | Robust background subtraction for network surveillance in H. 264 streaming video | |
| CN106851302B (en) | A kind of Moving Objects from Surveillance Video detection method based on intraframe coding compression domain | |
| US20150264357A1 (en) | Method and system for encoding digital images, corresponding apparatus and computer program product | |
| CN106534849B (en) | A kind of quick HEVC inter-frame encoding methods | |
| CN107005702B (en) | System and method for processing blocks of a digital image | |
| EP2782344A1 (en) | Inter-image prediction method and device and corresponding coding method and apparatus | |
| KR102261669B1 (en) | Artificial Neural Network Based Object Region Detection Method, Device and Computer Program Thereof | |
| Meuel et al. | Mesh-based piecewise planar motion compensation and optical flow clustering for ROI coding | |
| WO2016189404A1 (en) | Foreground motion detection in compressed video data | |
| CN104580825A (en) | Method for identifying and tracking object in video | |
| CN114157870A (en) | Encoding method, medium, and electronic device | |
| CN113422959A (en) | Video encoding and decoding method and device, electronic equipment and storage medium | |
| Moriyama et al. | Moving object detection in HEVC video by frame sub-sampling | |
| Lee et al. | Improve the HEVC algorithm complexity based on the visual perception | |
| CN103905818A (en) | Method for rapidly determining inter-frame prediction mode in HEVC standard based on Hough conversion | |
| Eleuch et al. | A study on the impact of multiview distributed feature coding on a multicamera vehicle tracking system at roundabouts | |
| CN110839157B (en) | Image processing method and device | |
| CN109862315B (en) | Video processing method, related device and computer storage medium | |
| CN105357494A (en) | Video encoding and decoding method and apparatus, and computer program product |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |