CN106021321B - The search method of online real-time video based on picture - Google Patents

The search method of online real-time video based on picture Download PDF

Info

Publication number
CN106021321B
CN106021321B CN201610294978.2A CN201610294978A CN106021321B CN 106021321 B CN106021321 B CN 106021321B CN 201610294978 A CN201610294978 A CN 201610294978A CN 106021321 B CN106021321 B CN 106021321B
Authority
CN
China
Prior art keywords
elements
binary
compression
gained
video
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.)
Active
Application number
CN201610294978.2A
Other languages
Chinese (zh)
Other versions
CN106021321A (en
Inventor
刘盾
杨丰羽
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Chengdu Sobey Digital Technology Co Ltd
Original Assignee
Chengdu Sobey Digital Technology Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Chengdu Sobey Digital Technology Co Ltd filed Critical Chengdu Sobey Digital Technology Co Ltd
Priority to CN201610294978.2A priority Critical patent/CN106021321B/en
Publication of CN106021321A publication Critical patent/CN106021321A/en
Application granted granted Critical
Publication of CN106021321B publication Critical patent/CN106021321B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • 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
    • G06F16/78—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
    • G06F16/783—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
    • G06F16/7844—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using original textual content or text extracted from visual content or transcript of audio data

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Library & Information Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Image Analysis (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The online real-time video search method based on picture that the invention discloses a kind of, to be suitable for the retrieval of online real-time video.It the steps include: that (1) server end extracts video stream characteristics data procedures, specially server end includes multiple video flowings in real time, nearest 5~1000 seconds videos in each video flowing are taken, characteristic of the video frequency feature data of each video as corresponding video flowing is extracted；(2) circulation executes step (1) and enters step (3) after the retrieval request that received server-side to client is initiated；The client initiates retrieval request process, and specially client obtains one or more picture, extracts the characteristic of one or more picture, and the characteristic of the picture is uploaded onto the server end；(3) retrieving specially takes the characteristic of picture obtained by step (2) and the characteristic of each video flowing obtained by step (1) to match；Subsequently back into step (1).

Description

The search method of online real-time video based on picture
Technical field
The search method of the present invention relates to a kind of online real-time video based on picture.
Background technique
Nearest more than ten years, computer vision technique rapid development, as David G. Lowe 2004 in Internation " the Distinctive Image Features from delivered on al Journal of Computer Vision Scale-Invariant Keypoints. ", so that the search of image content-based becomes practical, so that being based on video The video search of content become possible.Along with last decade, the computing capability of computer hardware equipment is improved and hard The decline of part cost and the appearance of various cloud computing platforms, if the cost of the E2C of Amazon, calculating are also greatly reduced, this makes Obtain the search based on video content also becomes practical gradually.The technical solution of mainstream uses for reference the side from image retrieval mostly at present Method, and relevant standard is had already appeared, for details, reference can be made to ISO/IEC DIS 15938-13.But current mainstream scheme is equal Consideration is offline video retrieval, and such scheme is first indexed all videos in video library, this index generally comprises Two parts: video frequency feature data and index feature data are extracted.Wherein content based video retrieval system, video frequency feature data Usually not only include local description, further includes global description's made of set local description, the calculation amount of both this is all It is not small；Meanwhile the characteristic index in the scheme of mainstream generally establishes MBIT according to global description's after binarization (Multi-block Index Table) can refer to the test model Test Model 13 of ISO/IEC JTC1/SC29: Compact Descriptors for Visual Search, but the foundation indexed can devote a tremendous amount of time.For offline The search of video, delay can be tolerated, but online real-time video is retrieved, to guarantee real-time, in order to cope with Big calculation amount, it will so that the cost of hardware investment is high；However since the application of Online Video retrieval is limited, Such as traffic camera search, station synchronization interaction etc., the quantity of live video stream is not especially big in these occasions, is built Lithol, which draws, can't bring too big performance boost.Therefore, video retrieval technology general at present is not particularly suited for online real-time The retrieval of video.
Summary of the invention
The inspection of it is an object of the invention to overcome the deficiencies of the prior art and provide a kind of online real-time video based on picture Suo Fangfa, this method omit Index process, without using global description's, real-time are effectively ensured, this method calculating is simple, is easy to It realizes.
The purpose of the present invention is achieved through the following technical solutions: the retrieval side of the online real-time video based on picture Method, it includes the following steps:
S1: multiple video flowings are included in server end extraction in real time, take nearest 5~1000 seconds views in each video flowing Frequently, characteristic of the video frequency feature data of each video as corresponding video flowing is extracted；
S2: circulation executes step S1, until receiving client retrieval request；
S3: client obtains one or more picture, calculates the binary features of picture, obtains according to picture successive Sequence, which uploads onto the server the binary features of picture, holds request retrieval；
S4: the binary features of client picture are matched with the binary features data of each video flowing, With value the maximum, if more than threshold value, then corresponding to video flowing is the video flowing to be retrieved, otherwise retrieval failure；
S5: output search result.
The extraction video frequency feature data includes following sub-step:
S11: a frame or the multiframe in video are taken；
S12: extracting the SIFT characteristic point of each frame in S11, multiple in selection gained SIFT feature, selected by calculating The Even-dimensional SIFT of the SIFT feature taken describes son；
S13: all Even-dimensional SIFT obtained by S12, which are described sub- boil down to, has the binary system of 16~512 binary digits special Sign.
The binary features of the calculating picture include following sub-step:
S31: extracting SIFT feature from picture, chooses multiple in gained SIFT feature, the selected SIFT spy of calculating The SIFT of sign point describes son；
S32: the binary features that sub- boil down to there are 16~512 binary digits are described into all SIFT obtained by S31.
It carries out also wanting record matching value while binary features Data Matching in the S4.
Described describes the binary features that sub- boil down to has 16~512 binary digits for Even-dimensional SIFT, specific side Method are as follows: the SIFT of N-dimensional, which describes son, can regard N number of number as, to N number of any combination of two of number, obtain N/2 several right, this N/2 Number is to arbitrary arrangement, and every number after arrangement is to corresponding bit, and wherein N is even number, and will be obtained by compressing by one There are the binary features of N/2 binary digit, each number is to a binary features are obtained, specially by each number pair First number and second number are compared, and first number is greater than second number, then compression obtain corresponding to the two of binary digit into Feature processed is 1；Less than second number of first number, then it is 0 that compression, which obtains the binary features of corresponding binary digit,.
The beneficial effects of the present invention are: the search method of the present invention provides a kind of online real-time video based on picture, Since without Index process, also without using global description's, real-time is guaranteed.New present invention further introduces one kind The method that son carries out binaryzation compression is described to SIFT, this method calculating is simple, is easily achieved, and the binary system after compressing Feature matches the Hamming distance also used and also achieves low calculation amount bring high efficiency while guaranteeing performance.
Detailed description of the invention
Fig. 1 is online real-time video searching method flow chart of the present invention；
Fig. 2 is embodiment system schematic；
Fig. 3 is that schematic diagram is numbered in 128 elements of 128 dimension SIFT description.
Specific embodiment
Technical solution of the present invention is described in further detail with reference to the accompanying drawing, but protection scope of the present invention is not limited to It is as described below.
Fig. 1 is online real-time video searching method flow chart of the present invention, and this example implements this method, with one or more Platform computer is server end, and using smart phone as client, server end and client are using same SIFT description son pressure Contracting method.Fig. 2 is the present embodiment system schematic.
Server end includes 10 video flowings in real time, extracts the characteristic of each video flowing, can implement as follows:
To each video flowing of input, handle in accordance with the following methods:
A. 20 seconds videos for taking video flowing nearest；
B. in above-mentioned resulting 20 seconds videos, every 0.5 second one frame of pumping；
C. each frame of releasing is handled as follows:
C-1. all SIFT features that the frame is extracted from frame, are ranked up according to response intensity, and it is strong to choose response Strongest 2000 SIFT features (then all choosing less than 2000) is spent, SIFT description of their 128 dimensions is calculated.
C-2. Fig. 3 is that 128 elements of each 128 dimension SIFT description are numbered respectively, uses following rule Each 128 dimension SIFT is described into one binary features for there are 64 binary digits of sub- boil down to and preservation；It is tieed up below to 128 The reference of element, the number referring to the signal in Fig. 3 in SIFT description:
If No. 0 element is greater than No. 80 elements, otherwise it is 0 that compression the 1st binary digit of gained, which is 1,；
If No. 1 element is greater than No. 81 elements, otherwise it is 0 that compression the 2nd binary digit of gained, which is 1,；
If No. 2 elements are greater than No. 82 elements, otherwise it is 0 that compression the 3rd binary digit of gained, which is 1,；
If No. 3 elements are greater than No. 83 elements, otherwise it is 0 that compression the 4th binary digit of gained, which is 1,；
If No. 4 elements are greater than No. 84 elements, otherwise it is 0 that compression the 5th binary digit of gained, which is 1,；
If No. 5 elements are greater than No. 85 elements, otherwise it is 0 that compression the 6th binary digit of gained, which is 1,；
If No. 6 elements are greater than No. 86 elements, otherwise it is 0 that compression the 7th binary digit of gained, which is 1,；
If No. 7 elements are greater than No. 87 elements, otherwise it is 0 that compression the 8th binary digit of gained, which is 1,；
If No. 40 elements are greater than No. 120 elements, otherwise it is 0 that compression the 9th binary digit of gained, which is 1,；
If No. 41 elements are greater than No. 121 elements, otherwise it is 0 that compression the 10th binary digit of gained, which is 1,；
If No. 42 elements are greater than No. 122 elements, otherwise it is 0 that compression the 11st binary digit of gained, which is 1,；
If No. 43 elements are greater than No. 123 elements, otherwise it is 0 that compression the 12nd binary digit of gained, which is 1,；
If No. 44 elements are greater than No. 124 elements, otherwise it is 0 that compression the 13rd binary digit of gained, which is 1,；
If No. 45 elements are greater than No. 125 elements, otherwise it is 0 that compression the 14th binary digit of gained, which is 1,；
If No. 46 elements are greater than No. 126 elements, otherwise it is 0 that compression the 15th binary digit of gained, which is 1,；
If No. 47 elements are greater than No. 127 elements, otherwise it is 0 that compression the 16th binary digit of gained, which is 1,；
If No. 24 elements are greater than No. 72 elements, otherwise it is 0 that compression the 17th binary digit of gained, which is 1,；
If No. 25 elements are greater than No. 73 elements, otherwise it is 0 that compression the 18th binary digit of gained, which is 1,；
If No. 26 elements are greater than No. 74 elements, otherwise it is 0 that compression the 19th binary digit of gained, which is 1,；
If No. 27 elements are greater than No. 75 elements, otherwise it is 0 that compression the 20th binary digit of gained, which is 1,；
If No. 28 elements are greater than No. 76 elements, otherwise it is 0 that compression the 21st binary digit of gained, which is 1,；
If No. 29 elements are greater than No. 77 elements, otherwise it is 0 that compression the 22nd binary digit of gained, which is 1,；
If No. 30 elements are greater than No. 78 elements, otherwise it is 0 that compression the 23rd binary digit of gained, which is 1,；
If No. 31 elements are greater than No. 79 elements, otherwise it is 0 that compression the 24th binary digit of gained, which is 1,；
If No. 48 elements are greater than No. 96 elements, otherwise it is 0 that compression the 25th binary digit of gained, which is 1,；
If No. 49 elements are greater than No. 97 elements, otherwise it is 0 that compression the 26th binary digit of gained, which is 1,；
If No. 50 elements are greater than No. 98 elements, otherwise it is 0 that compression the 27th binary digit of gained, which is 1,；
If No. 51 elements are greater than No. 99 elements, otherwise it is 0 that compression the 28th binary digit of gained, which is 1,；
If No. 52 elements are greater than No. 100 elements, otherwise it is 0 that compression the 29th binary digit of gained, which is 1,；
If No. 53 elements are greater than No. 101 elements, otherwise it is 0 that compression the 30th binary digit of gained, which is 1,；
If No. 54 elements are greater than No. 102 elements, otherwise it is 0 that compression the 31st binary digit of gained, which is 1,；
If No. 55 elements are greater than No. 103 elements, otherwise it is 0 that compression the 32nd binary digit of gained, which is 1,；
If No. 8 elements are greater than ununbium, otherwise it is 0 that compression the 33rd binary digit of gained, which is 1,；
If No. 9 elements are greater than ununtrium, otherwise it is 0 that compression the 34th binary digit of gained, which is 1,；
If No. 10 elements are greater than ununquadium, otherwise it is 0 that compression the 35th binary digit of gained, which is 1,；
If No. 11 elements are greater than ununpentium, otherwise it is 0 that compression the 36th binary digit of gained, which is 1,；
If No. 12 elements are greater than ununhexium, otherwise it is 0 that compression the 37th binary digit of gained, which is 1,；
If No. 13 elements are greater than ununseptium, otherwise it is 0 that compression the 38th binary digit of gained, which is 1,；
If No. 14 elements are greater than ununoctium, otherwise it is 0 that compression the 39th binary digit of gained, which is 1,；
If No. 15 elements are greater than ununennium, otherwise it is 0 that compression the 40th binary digit of gained, which is 1,；
If No. 16 elements are greater than No. 104 elements, otherwise it is 0 that compression the 41st binary digit of gained, which is 1,；
If No. 17 elements are greater than No. 105 elements, otherwise it is 0 that compression the 42nd binary digit of gained, which is 1,；
If No. 18 elements are greater than No. 106 elements, otherwise it is 0 that compression the 43rd binary digit of gained, which is 1,；
If No. 19 elements are greater than No. 107 elements, otherwise it is 0 that compression the 44th binary digit of gained, which is 1,；
If No. 20 elements are greater than No. 108 elements, otherwise it is 0 that compression the 45th binary digit of gained, which is 1,；
If No. 21 elements are greater than No. 109 elements, otherwise it is 0 that compression the 46th binary digit of gained, which is 1,；
If No. 22 elements are greater than ununnilium, otherwise it is 0 that compression the 47th binary digit of gained, which is 1,；
If No. 23 elements are greater than unununium, otherwise it is 0 that compression the 48th binary digit of gained, which is 1,；
If No. 32 elements are greater than No. 88 elements, otherwise it is 0 that compression the 49th binary digit of gained, which is 1,；
If No. 33 elements are greater than No. 89 elements, otherwise it is 0 that compression the 50th binary digit of gained, which is 1,；
If No. 34 elements are greater than No. 90 elements, otherwise it is 0 that compression the 51st binary digit of gained, which is 1,；
If No. 35 elements are greater than No. 91 elements, otherwise it is 0 that compression the 52nd binary digit of gained, which is 1,；
If No. 36 elements are greater than No. 92 elements, otherwise it is 0 that compression the 53rd binary digit of gained, which is 1,；
If No. 37 elements are greater than No. 93 elements, otherwise it is 0 that compression the 54th binary digit of gained, which is 1,；
If No. 38 elements are greater than No. 94 elements, otherwise it is 0 that compression the 55th binary digit of gained, which is 1,；
If No. 39 elements are greater than No. 95 elements, otherwise it is 0 that compression the 56th binary digit of gained, which is 1,；
If No. 64 elements are greater than No. 56 elements, otherwise it is 0 that compression the 57th binary digit of gained, which is 1,；
If No. 65 elements are greater than No. 57 elements, otherwise it is 0 that compression the 58th binary digit of gained, which is 1,；
If No. 66 elements are greater than No. 58 elements, otherwise it is 0 that compression the 59th binary digit of gained, which is 1,；
If No. 67 elements are greater than No. 59 elements, otherwise it is 0 that compression the 60th binary digit of gained, which is 1,；
If No. 68 elements are greater than No. 60 elements, otherwise it is 0 that compression the 61st binary digit of gained, which is 1,；
If No. 69 elements are greater than No. 61 elements, otherwise it is 0 that compression the 62nd binary digit of gained, which is 1,；
If No. 70 elements are greater than No. 62 elements, otherwise it is 0 that compression the 63rd binary digit of gained, which is 1,；
If No. 71 elements are greater than No. 63 elements, otherwise it is 0 that compression the 64th binary digit of gained, which is 1,；
D. the continuous A, B recycled above, step C, the retrieval request sent until receiving client.Receive retrieval request Afterwards, it is retrieved using such as under type:
D-1: all binary features for taking the client received to send, two obtained respectively with each frame of each stream System feature is matched, and when matching uses Hamming distance, if Hamming distance is less than 4, then it is assumed that two binary features matchings； The points that the binary features that each frame of all binary features and each stream that statistics client is sent obtains match.
D-2: the corresponding stream of that frame for taking matching points most is most matched stream；Most its match point of frame if matching is counted Number has been greater than 20, then retrieves success, and most matched stream is the stream to be retrieved, returns to search result and the successful message of retrieval to visitor Family end；Otherwise retrieval failure returns to retrieval failure news to client.
D-3: step A. is returned to
Cell phone client can be realized as follows:
Cell phone client persistently shoots picture, and handles each picture, which will carry out always, until clothes Business device, which returns, retrieves successful message, wherein being handled each picture, detailed process is as follows:
A. all SIFT features that the picture is extracted from each picture, are ranked up according to response intensity, choosing Strongest 2000 SIFT features of response intensity (then all choosing less than 2000) is taken, their 128 dimensions are calculated SIFT description.
B. Fig. 3 is that 128 elements of each 128 dimension SIFT description are numbered respectively, will using following rule Each 128 dimension SIFT describes one binary features for having 64 binary digits of sub- boil down to and preservation；It is tieed up below to 128 The reference of element, the number referring to the signal in Fig. 3 in SIFT description:
If No. 0 element is greater than No. 80 elements, otherwise it is 0 that compression the 1st binary digit of gained, which is 1,；
If No. 1 element is greater than No. 81 elements, otherwise it is 0 that compression the 2nd binary digit of gained, which is 1,；
If No. 2 elements are greater than No. 82 elements, otherwise it is 0 that compression the 3rd binary digit of gained, which is 1,；
If No. 3 elements are greater than No. 83 elements, otherwise it is 0 that compression the 4th binary digit of gained, which is 1,；
If No. 4 elements are greater than No. 84 elements, otherwise it is 0 that compression the 5th binary digit of gained, which is 1,；
If No. 5 elements are greater than No. 85 elements, otherwise it is 0 that compression the 6th binary digit of gained, which is 1,；
If No. 6 elements are greater than No. 86 elements, otherwise it is 0 that compression the 7th binary digit of gained, which is 1,；
If No. 7 elements are greater than No. 87 elements, otherwise it is 0 that compression the 8th binary digit of gained, which is 1,；
If No. 40 elements are greater than No. 120 elements, otherwise it is 0 that compression the 9th binary digit of gained, which is 1,；
If No. 41 elements are greater than No. 121 elements, otherwise it is 0 that compression the 10th binary digit of gained, which is 1,；
If No. 42 elements are greater than No. 122 elements, otherwise it is 0 that compression the 11st binary digit of gained, which is 1,；
If No. 43 elements are greater than No. 123 elements, otherwise it is 0 that compression the 12nd binary digit of gained, which is 1,；
If No. 44 elements are greater than No. 124 elements, otherwise it is 0 that compression the 13rd binary digit of gained, which is 1,；
If No. 45 elements are greater than No. 125 elements, otherwise it is 0 that compression the 14th binary digit of gained, which is 1,；
If No. 46 elements are greater than No. 126 elements, otherwise it is 0 that compression the 15th binary digit of gained, which is 1,；
If No. 47 elements are greater than No. 127 elements, otherwise it is 0 that compression the 16th binary digit of gained, which is 1,；
If No. 24 elements are greater than No. 72 elements, otherwise it is 0 that compression the 17th binary digit of gained, which is 1,；
If No. 25 elements are greater than No. 73 elements, otherwise it is 0 that compression the 18th binary digit of gained, which is 1,；
If No. 26 elements are greater than No. 74 elements, otherwise it is 0 that compression the 19th binary digit of gained, which is 1,；
If No. 27 elements are greater than No. 75 elements, otherwise it is 0 that compression the 20th binary digit of gained, which is 1,；
If No. 28 elements are greater than No. 76 elements, otherwise it is 0 that compression the 21st binary digit of gained, which is 1,；
If No. 29 elements are greater than No. 77 elements, otherwise it is 0 that compression the 22nd binary digit of gained, which is 1,；
If No. 30 elements are greater than No. 78 elements, otherwise it is 0 that compression the 23rd binary digit of gained, which is 1,；
If No. 31 elements are greater than No. 79 elements, otherwise it is 0 that compression the 24th binary digit of gained, which is 1,；
If No. 48 elements are greater than No. 96 elements, otherwise it is 0 that compression the 25th binary digit of gained, which is 1,；
If No. 49 elements are greater than No. 97 elements, otherwise it is 0 that compression the 26th binary digit of gained, which is 1,；
If No. 50 elements are greater than No. 98 elements, otherwise it is 0 that compression the 27th binary digit of gained, which is 1,；
If No. 51 elements are greater than No. 99 elements, otherwise it is 0 that compression the 28th binary digit of gained, which is 1,；
If No. 52 elements are greater than No. 100 elements, otherwise it is 0 that compression the 29th binary digit of gained, which is 1,；
If No. 53 elements are greater than No. 101 elements, otherwise it is 0 that compression the 30th binary digit of gained, which is 1,；
If No. 54 elements are greater than No. 102 elements, otherwise it is 0 that compression the 31st binary digit of gained, which is 1,；
If No. 55 elements are greater than No. 103 elements, otherwise it is 0 that compression the 32nd binary digit of gained, which is 1,；
If No. 8 elements are greater than ununbium, otherwise it is 0 that compression the 33rd binary digit of gained, which is 1,；
If No. 9 elements are greater than ununtrium, otherwise it is 0 that compression the 34th binary digit of gained, which is 1,；
If No. 10 elements are greater than ununquadium, otherwise it is 0 that compression the 35th binary digit of gained, which is 1,；
If No. 11 elements are greater than ununpentium, otherwise it is 0 that compression the 36th binary digit of gained, which is 1,；
If No. 12 elements are greater than ununhexium, otherwise it is 0 that compression the 37th binary digit of gained, which is 1,；
If No. 13 elements are greater than ununseptium, otherwise it is 0 that compression the 38th binary digit of gained, which is 1,；
If No. 14 elements are greater than ununoctium, otherwise it is 0 that compression the 39th binary digit of gained, which is 1,；
If No. 15 elements are greater than ununennium, otherwise it is 0 that compression the 40th binary digit of gained, which is 1,；
If No. 16 elements are greater than No. 104 elements, otherwise it is 0 that compression the 41st binary digit of gained, which is 1,；
If No. 17 elements are greater than No. 105 elements, otherwise it is 0 that compression the 42nd binary digit of gained, which is 1,；
If No. 18 elements are greater than No. 106 elements, otherwise it is 0 that compression the 43rd binary digit of gained, which is 1,；
If No. 19 elements are greater than No. 107 elements, otherwise it is 0 that compression the 44th binary digit of gained, which is 1,；
If No. 20 elements are greater than No. 108 elements, otherwise it is 0 that compression the 45th binary digit of gained, which is 1,；
If No. 21 elements are greater than No. 109 elements, otherwise it is 0 that compression the 46th binary digit of gained, which is 1,；
If No. 22 elements are greater than ununnilium, otherwise it is 0 that compression the 47th binary digit of gained, which is 1,；
If No. 23 elements are greater than unununium, otherwise it is 0 that compression the 48th binary digit of gained, which is 1,；
If No. 32 elements are greater than No. 88 elements, otherwise it is 0 that compression the 49th binary digit of gained, which is 1,；
If No. 33 elements are greater than No. 89 elements, otherwise it is 0 that compression the 50th binary digit of gained, which is 1,；
If No. 34 elements are greater than No. 90 elements, otherwise it is 0 that compression the 51st binary digit of gained, which is 1,；
If No. 35 elements are greater than No. 91 elements, otherwise it is 0 that compression the 52nd binary digit of gained, which is 1,；
If No. 36 elements are greater than No. 92 elements, otherwise it is 0 that compression the 53rd binary digit of gained, which is 1,；
If No. 37 elements are greater than No. 93 elements, otherwise it is 0 that compression the 54th binary digit of gained, which is 1,；
If No. 38 elements are greater than No. 94 elements, otherwise it is 0 that compression the 55th binary digit of gained, which is 1,；
If No. 39 elements are greater than No. 95 elements, otherwise it is 0 that compression the 56th binary digit of gained, which is 1,；
If No. 64 elements are greater than No. 56 elements, otherwise it is 0 that compression the 57th binary digit of gained, which is 1,；
If No. 65 elements are greater than No. 57 elements, otherwise it is 0 that compression the 58th binary digit of gained, which is 1,；
If No. 66 elements are greater than No. 58 elements, otherwise it is 0 that compression the 59th binary digit of gained, which is 1,；
If No. 67 elements are greater than No. 59 elements, otherwise it is 0 that compression the 60th binary digit of gained, which is 1,；
If No. 68 elements are greater than No. 60 elements, otherwise it is 0 that compression the 61st binary digit of gained, which is 1,；
If No. 69 elements are greater than No. 61 elements, otherwise it is 0 that compression the 62nd binary digit of gained, which is 1,；
If No. 70 elements are greater than No. 62 elements, otherwise it is 0 that compression the 63rd binary digit of gained, which is 1,；
If No. 71 elements are greater than No. 63 elements, otherwise it is 0 that compression the 64th binary digit of gained, which is 1,；
C. server end request inquiry is sent by the binary features that compression obtains.
By the description above it is found that relative to general video retrieval method, method of the invention is examined for Online Video This kind of special video frequency searching of rope is optimized.Since without Index process, also without using global description's, real-time is obtained Guarantee is arrived.Present invention further introduces a kind of new method for describing son progress binaryzation compression to SIFT, this method is calculated Simply, it is easily achieved, and the binary features after compressing, the Hamming distance also used is matched, while guaranteeing performance, Also achieve low calculation amount bring high efficiency.

Claims (2)

1.基于图片的在线实时视频的检索方法，其特征在于：它包括如下步骤：1. the retrieval method of the online real-time video based on picture, is characterized in that: it comprises the steps: S1：服务器端实时收录多个视频流，提取 所述每个视频流中最近5～1000秒的视频，提取所述每个视频的视频特征数据作为对应视频流的特征数据；S1: The server side records multiple video streams in real time, extracts the videos of the last 5 to 1000 seconds in each video stream, and extracts the video feature data of each video as the feature data of the corresponding video stream; S2：循环执行步骤S1，直到接收到客户端检索请求；S2: cyclically execute step S1 until a client retrieval request is received; S3：客户端获取一张或多张图片，计算出图片的二进制特征，按照图片获取的先后顺序将图片的二进制特征上传到服务器端请求检索；S3: The client obtains one or more pictures, calculates the binary features of the pictures, and uploads the binary features of the pictures to the server for retrieval according to the order of picture acquisition; S4：将客户端图片的二进制特征与每一个视频流的二进制特征数据进行匹配，匹配值最大者，若大于了阈值，则对应视频流为要检索的视频流，否则检索失败；S4: Match the binary feature of the client picture with the binary feature data of each video stream, the one with the largest matching value, if it is greater than the threshold, the corresponding video stream is the video stream to be retrieved, otherwise the retrieval fails; S5：输出检索结果；S5: output the retrieval result; 步骤S1所述的提取视频特征数据包括如下子步骤：The extraction of video feature data described in step S1 includes the following sub-steps: S11：取视频中的一帧或多帧；S11: take one or more frames in the video; S12：提取S11中每一帧的 SIFT 特征点，选取所得SIFT特征点中的多个，计算所选取的SIFT特征点的偶数维SIFT描述子；S12: extract the SIFT feature points of each frame in S11, select a plurality of the obtained SIFT feature points, and calculate the even-dimensional SIFT descriptor of the selected SIFT feature points; S13：将S12所得所有偶数维SIFT描述子压缩为有16～512个二进制位的二进制特征；S13: compress all even-dimensional SIFT descriptors obtained in S12 into binary features with 16-512 binary bits; 步骤S3所述的计算图片的二进制特征包括如下子步骤：The calculation of the binary feature of the picture described in step S3 includes the following sub-steps: S31：从图片中提取SIFT特征点，选取所得SIFT特征点中的多个，计算所选SIFT特征点的SIFT描述子；S31: Extract SIFT feature points from the picture, select a plurality of the obtained SIFT feature points, and calculate the SIFT descriptor of the selected SIFT feature points; S32：将S31所得所有SIFT描述子压缩为有16～512个二进制位的二进制特征；S32: compress all SIFT descriptors obtained in S31 into binary features with 16 to 512 binary bits; 所述的将偶数维SIFT描述子压缩为有16～512个二进制位的二进制特征，具体方法为：N维的SIFT描述子可看成N个数，对所述N个数任意两两组合，得到 N/2 个数对，这N/2个数对任意排列，排列后的每个数对对应一个二进制位，其中N为偶数，而压缩将得到一个有N/2个二进制位的二进制特征,每一个数对得到一个二进制特征，具体为将每一个数对的第一个数和第二个数进行比较，第一个数大于第二个数，则压缩得到对应二进制位的二进制特征为1；第一个数小于第二个数，则压缩得到对应二进制位的二进制特征为0。The even-dimensional SIFT descriptor is compressed into binary features with 16 to 512 binary bits, and the specific method is as follows: the N-dimensional SIFT descriptor can be regarded as N numbers, and the N numbers are arbitrarily combined in pairs, Get N/2 number pairs, these N/2 number pairs are arranged arbitrarily, each number pair after arrangement corresponds to a binary bit, where N is an even number, and compression will get a binary feature with N/2 binary bits , Each pair of numbers gets a binary feature, specifically by comparing the first number and the second number of each number pair, the first number is greater than the second number, then the binary feature of the corresponding binary bit is obtained by compressing 1; if the first number is less than the second number, the binary characteristic of the corresponding binary bit is 0. 2.根据权利要求1所述的基于图片的在线实时视频的检索方法，其特征在于：所述的S4中进行二进制特征数据匹配的同时还要记录匹配值。2. The retrieval method of picture-based online real-time video according to claim 1, is characterized in that: in the described S4, the matching value is also recorded while the binary feature data matching is performed.
CN201610294978.2A 2016-05-06 2016-05-06 The search method of online real-time video based on picture Active CN106021321B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610294978.2A CN106021321B (en) 2016-05-06 2016-05-06 The search method of online real-time video based on picture

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610294978.2A CN106021321B (en) 2016-05-06 2016-05-06 The search method of online real-time video based on picture

Publications (2)

Publication Number Publication Date
CN106021321A CN106021321A (en) 2016-10-12
CN106021321B true CN106021321B (en) 2019-05-21

Family

ID=57082107

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610294978.2A Active CN106021321B (en) 2016-05-06 2016-05-06 The search method of online real-time video based on picture

Country Status (1)

Country Link
CN (1) CN106021321B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108509458B (en) * 2017-02-28 2022-12-16 阿里巴巴集团控股有限公司 Business object identification method and device
CN111506772B (en) * 2019-01-31 2023-05-09 北京小犀智能科技中心(有限合伙) Image searching method and system based on image feature extraction

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6021231A (en) * 1993-10-20 2000-02-01 Hitachi, Ltd. Video retrieval method and apparatus
WO2007073053A1 (en) * 2005-12-20 2007-06-28 Nhn Corporation Method for searching situation of moving picture and system using the same
CN101339662A (en) * 2008-08-13 2009-01-07 北京六维世纪网络技术有限公司 Method and device for creating video frequency feature data
CN101394522A (en) * 2007-09-19 2009-03-25 中国科学院计算技术研究所 Method and system for detecting video copy
CN102103609A (en) * 2009-12-21 2011-06-22 北京中星微电子有限公司 Information retrieval method and system
CN102117313A (en) * 2010-12-29 2011-07-06 天脉聚源(北京)传媒科技有限公司 Video retrieval method and system
CN102760169A (en) * 2012-06-13 2012-10-31 天脉聚源(北京)传媒科技有限公司 Method for detecting advertising slots in television direct transmission streams
CN103136292A (en) * 2011-12-05 2013-06-05 中国移动通信集团公司 Image retrieval method and image retrieval device based on image characteristics
CN103902569A (en) * 2012-12-27 2014-07-02 重庆凯泽科技有限公司 Video matching method based on Bag of Words

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6021231A (en) * 1993-10-20 2000-02-01 Hitachi, Ltd. Video retrieval method and apparatus
WO2007073053A1 (en) * 2005-12-20 2007-06-28 Nhn Corporation Method for searching situation of moving picture and system using the same
CN101394522A (en) * 2007-09-19 2009-03-25 中国科学院计算技术研究所 Method and system for detecting video copy
CN101339662A (en) * 2008-08-13 2009-01-07 北京六维世纪网络技术有限公司 Method and device for creating video frequency feature data
CN102103609A (en) * 2009-12-21 2011-06-22 北京中星微电子有限公司 Information retrieval method and system
CN102117313A (en) * 2010-12-29 2011-07-06 天脉聚源(北京)传媒科技有限公司 Video retrieval method and system
CN103136292A (en) * 2011-12-05 2013-06-05 中国移动通信集团公司 Image retrieval method and image retrieval device based on image characteristics
CN102760169A (en) * 2012-06-13 2012-10-31 天脉聚源(北京)传媒科技有限公司 Method for detecting advertising slots in television direct transmission streams
CN103902569A (en) * 2012-12-27 2014-07-02 重庆凯泽科技有限公司 Video matching method based on Bag of Words

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
基于HBase的海量小视频存储与检索系统的研究与实现;刘晓静;《中国优秀硕士学位论文全文数据库 信息科技辑》;20141115(第11期);全文

Also Published As

Publication number Publication date
CN106021321A (en) 2016-10-12

Similar Documents

Publication Publication Date Title
Hu et al. Learning supervised scoring ensemble for emotion recognition in the wild
Kumar et al. Equal partition based clustering approach for event summarization in videos
CN100493195C (en) A Time-Shifted TV Video Matching Method Combining Program Content Metadata and Content Analysis
Chary et al. Feature extraction methods for color image similarity
CN105744292B (en) A kind of processing method and processing device of video data
CN101720006B (en) Positioning method suitable for representative frame extracted by video keyframe
Thomas et al. Perceptual video summarization—A new framework for video summarization
CN106021321B (en) The search method of online real-time video based on picture
Garcia et al. Dress like a star: Retrieving fashion products from videos
CN107527010B (en) Method for extracting video gene according to local feature and motion vector
CN104486585A (en) Method and system for managing urban mass surveillance video based on GIS
CN113743352A (en) Method and device for comparing similarity of video contents
CN108960041B (en) Image feature extraction method and device
CN112070047B (en) Video processing method, device and electronic equipment
Mademlis et al. Summarization of human activity videos using a salient dictionary
Yao et al. Object based video synopsis
CN106021320B (en) The search method of online real-time video based on video
CN112818863A (en) Method and system for constructing personnel track based on face searching
CN108764258B (en) Optimal image set selection method for group image insertion
KR100876280B1 (en) Statistical Shape Descriptor Extraction Apparatus and Method and Its Video Indexing System
WO2013018913A1 (en) Video processing system, method of determining viewer preference, video processing apparatus, and control method and control program therefor
Thepade et al. Extraction of key frames from video using discrete cosine transform
Xin et al. Robust feature selection with self-matching score
Liu et al. Within and between shot information utilisation in video key frame extraction
CN104850600A (en) Method and device for searching images containing faces

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