CN105894541B - A kind of moving target search method and system based on the collision of more videos - Google Patents
A kind of moving target search method and system based on the collision of more videos Download PDFInfo
- Publication number
- CN105894541B CN105894541B CN201610239606.XA CN201610239606A CN105894541B CN 105894541 B CN105894541 B CN 105894541B CN 201610239606 A CN201610239606 A CN 201610239606A CN 105894541 B CN105894541 B CN 105894541B
- Authority
- CN
- China
- Prior art keywords
- moving target
- video
- collision
- matching
- videos
- 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
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention relates to a kind of moving target search method and system based on the collision of more videos, method is the following steps are included: S1, each video difference extraction moving target snapshot to participation collision；S2, its characteristic value is extracted to the moving target snapshot of each video；S3, it participates in being collided two-by-two between the video collided, obtains the moving target snapshot for meeting similarity requirement, and form object matching to sequence；S4, all object matchings analyze sequence, finally retrieves the same movement target occurred in participating in collision video.Identical moving target in two videos can be fast and efficiently found out, improved efficiency is clearly.
Description
Technical field
The present invention relates to video investigation fields, and more specifically, it relates to a kind of moving targets based on the collision of more videos
Search method and system.
Background technique
Moving target, including people, vehicle, object are the core informations in monitor video.And what searching occurred in two videos
Same movement target, it is significant for cracking of cases, be conducive to quick lock in suspect；
And this work at present is always manually completed, and it is very time-consuming and laborious and inefficient, it is easy to miss important
Information is the important hindering factor for restricting case and quickly tracking down.Although having the technology for extracting the snapshot of moving target in video
And application software, but may often have thousands of snapshots in each video, it, be more in the case where no priori conditions
Identical moving target or a very intractable thing are looked in a video.
Summary of the invention
In view of this, it is necessary in view of the above-mentioned problems, providing a kind of moving target search method based on the collision of more videos
And system, identical moving target in two videos can be fast and efficiently found out, improved efficiency is clearly.
To achieve the goals above, technical scheme is as follows:
A kind of moving target search method based on the collision of more videos, comprising the following steps:
S1, moving target snapshot is extracted respectively to each video for participating in collision；
S2, its characteristic value is extracted to the moving target snapshot of each video；
S3, it participates in being collided two-by-two between the video collided, obtains the moving target snapshot for meeting similarity requirement, and
Object matching is formed to sequence；
S4, all object matchings analyze sequence, finally retrieves the phase occurred in participating in collision video
Same moving target.
Preferably, the step S3 is specifically included:
S301, the feature vector for calculating two video frequency motion target snapshots, the moving target snapshot of the every two video
It is matched two-by-two, forms moving target matching pair, the Euclidean distance calculating the feature vector of moving target matching between；
S302, the Euclidean distance of the feature vector of moving target matching pair is converted into similarity value between [0,1]；
S303, the suitable similarity threshold of setting, filtering exclude the moving target matching pair that similarity is less than threshold value；
S304, filtered moving target is matched to progress descending arrangement, and exports ranking results.
Preferably, the moving target snapshot of the every two video is matched two-by-two, it will be in one of video
Each moving target snapshot matched respectively with each moving target snapshot in another video, calculate target signature to
The Euclidean distance of amount shares M*N object matching pair, and wherein M, N are respectively the corresponding moving target snapshot number of two videos.
Preferably, in the step S302, similarity value=(1- feature vector Euclidean distance/fixed value).
Preferably, the similarity threshold is 0.65, the moving target to similarity greater than 0.65 is arranged into descending
Column, and exported.
A kind of moving target searching system based on the collision of more videos, including video acquisition module, characteristic extracting module, touch
Hit matching module, data server；
The video acquisition module is used to receive the video that acquisition participates in collision；
The characteristic extracting module is used to extract moving target snapshot respectively to the video for participating in collision, and extracts movement mesh
Mark the characteristic value of snapshot；
The collision matching module extracts the moving target for meeting similarity requirement for being collided two-by-two to video
Snapshot forms object matching to sequence；
For analyzing sequence all object matchings, retrieval obtains appearing in participation collision the data server
The same movement target occurred in video.
Preferably, further including a data memory module, for storing the video data received and search result.
Preferably, the collision matching module is carried out two-by-two by moving target snapshot to every two video
Match, form moving target matching pair, the Euclidean distance calculating the feature vector of moving target matching between converts Euclidean distance
Similarity value and screening and sequencing is carried out between [0,1], the final object matching that forms is to sequence.
Preferably, the collision matching module further includes excluding similarity to suitable threshold value is set and being less than threshold value
Moving target matching pair, to filtered moving target matching to carry out descending arrangement.
Preferably, the threshold value is set as 0.65.
Compared with prior art, the beneficial effects of the present invention are: the present invention passes through to the movement mesh for stating every two video
Mark snapshot matched two-by-two, formed moving target matching pair, calculate moving target match between feature vector Euclidean away from
From Euclidean distance is converted to similarity value and carries out screening and sequencing, the final object matching that forms is to sequence, to all targets
Matched sequence is analyzed, and the same movement target occurred in participating in collision video is finally retrieved, can quickly, efficiently
Identical moving target in two videos is found out on ground, and improved efficiency is clearly.
Detailed description of the invention
Fig. 1 is flow chart of the method for the present invention；
Fig. 2 is the specific flow chart of step S3 in the method for the present invention；
Fig. 3 is system structure diagram of the invention.
Specific embodiment
With reference to the accompanying drawings and examples to a kind of moving target retrieval side based on the collision of more videos of the present invention
Method and system are described further.
It is a kind of best reality of moving target search method and system based on the collision of more videos of the present invention below
Therefore example does not limit the scope of protection of the present invention.
A kind of moving target search method based on the collision of more videos, comprising the following steps:
S1, moving target snapshot is extracted respectively to each video for participating in collision；
S2, its characteristic value is extracted to the moving target snapshot of each video；
S3, it participates in being collided two-by-two between the video collided, obtains the moving target snapshot for meeting similarity requirement, and
Object matching is formed to sequence；
S4, all object matchings analyze sequence, finally retrieves the phase occurred in participating in collision video
Same moving target.
Preferably, the step S3 is specifically included:
S301, the feature vector for calculating two video frequency motion target snapshots, the moving target snapshot of the every two video
It is matched two-by-two, forms moving target matching pair, the Euclidean distance calculating the feature vector of moving target matching between；
S302, the Euclidean distance of the feature vector of moving target matching pair is converted into similarity value between [0,1]；
S303, the suitable similarity threshold of setting, filtering exclude the moving target matching pair that similarity is less than threshold value；
S304, filtered moving target is matched to progress descending arrangement, and exports ranking results.
Preferably, the moving target snapshot of the every two video is matched two-by-two, it will be in one of video
Each moving target snapshot matched respectively with each moving target snapshot in another video, calculate target signature to
The Euclidean distance of amount shares M*N object matching pair, and wherein M, N are respectively the corresponding moving target snapshot number of two videos.
Assuming that there are 3 videos to participate in crash analysis, then allows video 1 to be collided with video 2 respectively, handle and obtain one group
Object matching is to sequence；Video 2 is collided with video 3, is handled and is obtained one group of object matching sequence；Video 3 and video 1 into
Row collision, handles and obtains one group of object matching to sequence.
Preferably, in the step S302, similarity value=(1- feature vector Euclidean distance/fixed value).It can be with
Obtain the similarity value of moving target pair.
Preferably, the similarity threshold is 0.65, the moving target to similarity greater than 0.65 is arranged into descending
Column, and exported.
All object matchings analyze sequence, finally retrieve the identical fortune occurred in participating in collision video
Moving-target specifically includes:
Assuming that the object matching of video 1 and video 2 is to having object matching to [10,20], i.e. the 10th in video 1 in sequence
A target snapshot and the 20th snapshot similarity in video 2 are higher；The object matching sequence centering of video 2 and video 3 simultaneously
There is object matching to [20,15]；The object matching of video 1 and video 3, then can be with to having object matching in sequence to [10,15]
Determine [10,20,15] to the higher object matching pair of similarity, i.e. the 10th in video 1 target snapshot, the in video 2
The similarity of the 15th target snapshot in 20 target snapshots, video 3 is higher, may be same target.
A kind of moving target searching system based on the collision of more videos, including video acquisition module, characteristic extracting module, touch
Hit matching module, data server；
The video acquisition module is used to receive the video that acquisition participates in collision；
The characteristic extracting module is used to extract moving target snapshot respectively to the video for participating in collision, and extracts movement mesh
Mark the characteristic value of snapshot；
The collision matching module extracts the moving target for meeting similarity requirement for being collided two-by-two to video
Snapshot forms object matching to sequence；
For analyzing sequence all object matchings, retrieval obtains appearing in participation collision the data server
The same movement target occurred in video.
Preferably, further including a data memory module, for storing the video data received and search result.
Preferably, the collision matching module is carried out two-by-two by moving target snapshot to every two video
Match, form moving target matching pair, the Euclidean distance calculating the feature vector of moving target matching between converts Euclidean distance
Similarity value and screening and sequencing is carried out between [0,1], the final object matching that forms is to sequence.
Preferably, the collision matching module further includes excluding similarity to suitable threshold value is set and being less than threshold value
Moving target matching pair, to filtered moving target matching to carry out descending arrangement.
Preferably, the threshold value is set as 0.65.Moving target to similarity greater than 0.65 is to progress descending row
Sequence, and being exported, is conducive to that user is more convenient to quickly find similar purpose, so as to exclude a large amount of different fortune
Moving-target matching pair, is effectively reduced the time-consuming of storage, sequence.
As stated above, the present invention forms movement by being matched two-by-two to the moving target snapshot for stating every two video
Object matching pair, the Euclidean distance calculating the feature vector of moving target matching between, is converted to similarity value for Euclidean distance
And screening and sequencing is carried out, the final object matching that forms analyzes sequence all object matchings sequence, final to retrieve
The same movement target occurred in participating in collision video out, can fast and efficiently find out identical movement in two videos
Target, improved efficiency is clearly.
The embodiments described above only express several embodiments of the present invention, and the description thereof is more specific and detailed, but simultaneously
Limitations on the scope of the patent of the present invention therefore cannot be interpreted as.It should be pointed out that for those of ordinary skill in the art
For, without departing from the inventive concept of the premise, various modifications and improvements can be made, these belong to guarantor of the invention
Protect range.Therefore, the scope of protection of the patent of the invention shall be subject to the appended claims.
Claims (8)
1. a kind of moving target search method based on the collision of more videos, which comprises the following steps:
S1, moving target snapshot is extracted respectively to each video for participating in collision；
S2, its characteristic value is extracted to the moving target snapshot of each video；
S3, it participates in being collided two-by-two between the video collided, obtains the moving target snapshot for meeting similarity requirement, and form
Object matching is to sequence；
S4, all object matchings analyze sequence, finally retrieves the identical fortune occurred in participating in collision video
Moving-target；
The step S3 is specifically included:
The moving target snapshot of S301, the feature vector for calculating two video frequency motion target snapshots, every two video carry out two-by-two
Matching forms moving target matching pair, the Euclidean distance calculating the feature vector of moving target matching between；
S302, the Euclidean distance of the feature vector of moving target matching pair is converted into similarity value between [0,1]；
S303, the suitable similarity threshold of setting, filtering exclude the moving target matching pair that similarity is less than threshold value；
S304, filtered moving target is matched to progress descending arrangement, and exports ranking results.
2. the moving target search method according to claim 1 based on the collision of more videos, which is characterized in that described every two
The moving target snapshot of a video is matched two-by-two, by each moving target snapshot in one of video respectively with it is another
Each moving target snapshot in a video is matched, and the Euclidean distance of target feature vector is calculated, and shares M*N target
Pairing, wherein M, N are respectively the corresponding moving target snapshot number of two videos.
3. the moving target search method according to claim 1 based on the collision of more videos, which is characterized in that the step
In S302, similarity value=(1- feature vector Euclidean distance/fixed value).
4. the moving target search method according to claim 1 based on the collision of more videos, which is characterized in that described similar
Spending threshold value is 0.65, and the moving target to similarity greater than 0.65 is arranged into descending, and is exported.
5. a kind of moving target searching system based on the collision of more videos, which is characterized in that mentioned including video acquisition module, feature
Modulus block, collision matching module, data server；
The video acquisition module is used to receive the video that acquisition participates in collision；
The characteristic extracting module is used to extract moving target snapshot respectively to the video for participating in collision, and it is fast to extract moving target
According to characteristic value；
The collision matching module extracts that meet the moving target that similarity requires fast for being collided two-by-two to video
According to composition object matching is to sequence；
For analyzing sequence all object matchings, retrieval obtains appearing in participation collision video the data server
The same movement target of middle appearance；
The collision matching module is matched two-by-two by the moving target snapshot to every two video, forms moving target
Pairing, the Euclidean distance calculating the feature vector of moving target matching between, is converted to similarity between [0,1] for Euclidean distance
It is worth and carries out screening and sequencing, the final object matching that forms is to sequence.
6. the moving target searching system according to claim 5 based on the collision of more videos, which is characterized in that further include one
Data memory module, for storing the video data received and search result.
7. the moving target searching system according to claim 5 based on the collision of more videos, which is characterized in that the collision
Matching module further includes the moving target matching pair that similarity is less than threshold value being excluded, to filtered to suitable threshold value is set
Moving target matching is to progress descending arrangement.
8. the moving target searching system according to claim 7 based on the collision of more videos, which is characterized in that the threshold value
It is set as 0.65.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610239606.XA CN105894541B (en) | 2016-04-18 | 2016-04-18 | A kind of moving target search method and system based on the collision of more videos |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610239606.XA CN105894541B (en) | 2016-04-18 | 2016-04-18 | A kind of moving target search method and system based on the collision of more videos |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105894541A CN105894541A (en) | 2016-08-24 |
| CN105894541B true CN105894541B (en) | 2019-05-17 |
Family
ID=56705100
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610239606.XA Active CN105894541B (en) | 2016-04-18 | 2016-04-18 | A kind of moving target search method and system based on the collision of more videos |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105894541B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107622667B (en) * | 2017-09-28 | 2021-05-25 | 广州烽火众智数字技术有限公司 | Method and system for detecting altered license plate number |
| CN110263830B (en) * | 2019-06-06 | 2021-06-08 | 北京旷视科技有限公司 | Image processing method, device and system and storage medium |
| CN113420183B (en) * | 2021-06-21 | 2022-12-02 | 成都神州数码索贝科技有限公司 | Cross-video target retrieval and collision method and system in video detection system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102222104A (en) * | 2011-06-23 | 2011-10-19 | 华南理工大学 | Method for intelligently extracting video abstract based on time-space fusion |
| CN102598113A (en) * | 2009-06-30 | 2012-07-18 | 安芯美特控股有限公司 | Method circuit and system for matching an object or person present within two or more images |
| CN104636497A (en) * | 2015-03-05 | 2015-05-20 | 四川智羽软件有限公司 | Intelligent video data retrieval method |
| CN105335469A (en) * | 2015-09-29 | 2016-02-17 | 武汉传神信息技术有限公司 | Method and device for image matching and retrieving |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| MX353871B (en) * | 2014-02-10 | 2018-02-01 | Geenee Gmbh | Systems and methods for image-feature-based recognition. |
-
2016
- 2016-04-18 CN CN201610239606.XA patent/CN105894541B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102598113A (en) * | 2009-06-30 | 2012-07-18 | 安芯美特控股有限公司 | Method circuit and system for matching an object or person present within two or more images |
| CN102222104A (en) * | 2011-06-23 | 2011-10-19 | 华南理工大学 | Method for intelligently extracting video abstract based on time-space fusion |
| CN104636497A (en) * | 2015-03-05 | 2015-05-20 | 四川智羽软件有限公司 | Intelligent video data retrieval method |
| CN105335469A (en) * | 2015-09-29 | 2016-02-17 | 武汉传神信息技术有限公司 | Method and device for image matching and retrieving |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105894541A (en) | 2016-08-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Bhat et al. | Unveiling the power of deep tracking | |
| Ranjan et al. | Iterative crowd counting | |
| Zheng et al. | Person re-identification meets image search | |
| CN107169106B (en) | Video retrieval method, device, storage medium and processor | |
| CN105184238A (en) | Human face recognition method and system | |
| CN105894541B (en) | A kind of moving target search method and system based on the collision of more videos | |
| CN103150375A (en) | Quick video retrieval system and quick video retrieval method for video detection | |
| CN103207898A (en) | Method for rapidly retrieving similar faces based on locality sensitive hashing | |
| Ha et al. | Multiple object tracking using SIFT features and location matching | |
| CN103955518A (en) | Method for matching a test object spectrum with database spectrum | |
| CN103325122A (en) | Pedestrian retrieval method based on bidirectional sequencing | |
| CN104573681A (en) | Face recognition method | |
| CN103440508A (en) | Remote sensing image target recognition method based on visual word bag model | |
| US20200342023A1 (en) | Hash-based media search | |
| CN106933861A (en) | A kind of customized across camera lens target retrieval method of supported feature | |
| CN104881640A (en) | Method and device for acquiring vectors | |
| CN106127235B (en) | Vehicle query method and system based on target feature collision | |
| Kopp et al. | Interpreting and clustering outliers with sapling random forests | |
| Rao et al. | Using machines to improve human saliency detection | |
| Zhao et al. | Person re-identification with effectively designed parts | |
| Vezzani et al. | Benchmarking for Person Re-identification | |
| CN104143089B (en) | The critical point detection method based on space-time Energy Decomposition in human action identification | |
| Yang et al. | Multiple scaled person re-identification framework for HD Video surveillance application | |
| Wang et al. | Moving Target Classification Technique in Video Sequence | |
| Chang et al. | Manipulation classification for near-duplicate videos |
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 |