CN110276779A - A kind of dense population image generating method based on the segmentation of front and back scape - Google Patents
A kind of dense population image generating method based on the segmentation of front and back scape Download PDFInfo
- Publication number
- CN110276779A CN110276779A CN201910480368.5A CN201910480368A CN110276779A CN 110276779 A CN110276779 A CN 110276779A CN 201910480368 A CN201910480368 A CN 201910480368A CN 110276779 A CN110276779 A CN 110276779A
- Authority
- CN
- China
- Prior art keywords
- image
- scene
- crowd
- dense population
- back scape
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
- G06V20/53—Recognition of crowd images, e.g. recognition of crowd congestion
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30196—Human being; Person
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Library & Information Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Processing Or Creating Images (AREA)
Abstract
The invention discloses a kind of dense population image generating methods based on the segmentation of front and back scape, its main feature is that in using front and back scape segmentation network model and vanishing Point Detection Method search to have an X-rayed similar method, crowd's (prospect) that cutting obtains is pieced together with new scene image, the dense population image for generating more scenes, the express statistic for crowd's quantity under the scene.The present invention has simple, easy compared with prior art, accuracy rate is high, utilize the data for there are character positions to mark, divide personage's (prospect) and the background of network segmented image by front and back scape, similar perspective is found by way of search again and pieces the new data of generation together from scene figure, compared to the true picture for manually marking new scene multiplicity is allowed again, required cost is smaller, and more true.
Description
Technical field
The present invention relates to crowd's counting technology field, especially a kind of image generating method based on the segmentation of front and back scape.
Background technique
As number increases, the activity that people intensively gather increases, and brings completely new security monitoring, proactive problem, thus
Designing a kind of the method for crowd's quantity can become particularly significant under express statistic scene.Computer can be by neural network
Study, express statistic one opens in image there are how many people, but the realization of the function needs first to instruct neural network model
Practice, which needs the support of the existing image for accurately marking independent character positions.It is current existing for this generic task
Image resource only contains the image obtained in a small number of several scenes, is not enough to that neural network model is allowed to have good scene
The scene changes of adaptability, i.e., existing marked image data are insufficient, cause current neural network model without calligraphy learning to enough
Scene resolving ability cannot show good performance in new application environment.
There is provided the changeable data set of more scenes for neural network model will help model to improve performance, but for the crowd
Counting load, training need to provide the specific coordinate position of each of figure with image.The label of the position needs artificial mark
Note, current existing image set, usually contains several hundred images, single image personage's reference numerals can achieve thousands of people, to this
Labor workload it is very big, but still be far from satisfying the needs of trained neural network.The workload is very big, cause using
Same method generates new data set very expensive, it is difficult to realize, simultaneously for the neural network mould of crowd's counting load design
Type, it is also not enough to the recognition capability of scene, cause performance to be difficult to satisfactory, so needing one kind at present can quickly generate
A large amount of new images that can be used for the task generate method.
Summary of the invention
The purpose of the present invention is in view of the deficiencies of the prior art and design it is a kind of based on front and back scape segmentation dense population
Image generating method, using front and back scape cutting techniques, crowd's (prospect) and background in separate picture, using currently having had just
Crowd's enumeration data collection image of true character positions mark, obtains crowd therein (prospect) for cutting and pieces other a variety of fields together
In scape image, the dense population image of scene multiplicity is generated, realizes the express statistic of crowd's quantity under the scene, using having had
The data of character positions mark, personage's (prospect) and the background of network segmented image are divided by front and back scape, then pass through searcher
Formula, which finds similar perspective, to be pieced together from scene figure and generates new data, compared to allowing the true figure for manually marking new scene multiplicity again
Picture, it is simple, easy, not only count at low cost, and more true, it is with a high credibility.
The object of the present invention is achieved like this: a kind of dense population image generating method based on the segmentation of front and back scape,
Feature is to have an X-rayed similar method using front and back scape segmentation network model and vanishing Point Detection Method search, the crowd that cutting is obtained
(prospect) is pieced together with new scene image, generates the dense population image of more scenes, realizes the quick of crowd's quantity under the scene
Statistics, detailed process the following steps are included:
A step: the dense population image that will acquire utilizes front and back scape segmentation network model, isolates crowd's (prospect) and background,
Crowd's picture set is obtained, the dense population image is existing true intensive scene image；The background image is not wrap
Various scene pictures containing personage.
B step: using the similar image search method of end point, and finding has similar perspective view to background image
Scene is as the alternate scenes picture pieced together.
Step c: crowd's picture and the alternate scenes picture in b step that a step obtains are pieced together, more scenes are generated
Dense population image, for crowd's quantity under the express statistic scene.
The present invention has simple, easy compared with prior art, and accuracy rate is high, utilizes the number for having had character positions to mark
According to, by front and back scape divide network segmented image personage's (prospect) and background, then by way of search find it is similar have an X-rayed from
Scene figure, which is pieced together, generates new data, and compared to the true picture for manually marking new scene multiplicity is allowed again, required cost is smaller,
And it is more true.
Detailed description of the invention
Fig. 1 is flow diagram of the invention；
Fig. 2 is specifically to use schematic diagram.
Specific embodiment
Refering to attached drawing 1, the present invention uses front and back scape segmentation network model and vanishing Point Detection Method search to have an X-rayed similar method,
Crowd's (prospect) that cutting obtains is pieced together with new scene image, the dense population image of more scenes is generated, realizes the scene
The express statistic of lower crowd's quantity, detailed process the following steps are included:
A step: the dense population image that will acquire utilizes front and back scape segmentation network model, isolates crowd's (prospect) and background,
Crowd's picture set is obtained, the dense population image is existing true intensive scene image；The background image is not wrap
Various scene pictures containing personage.
B step: using the similar image search method of end point, and finding has similar perspective view to background image
Scene is as the alternate scenes picture pieced together.
Step c: crowd's picture and the alternate scenes picture in b step that a step obtains are pieced together, more scenes are generated
Dense population image, for crowd's quantity under the express statistic scene.
The present invention is described in further detail for following mask body implementation.
Embodiment 1
Refering to attached drawing 2, scape segmentation network model, crowd's (prospect) of segmented image and background before and after the present invention utilizes, then by
Vanishing Point Detection Method search to original image there is the environment candidate of similar perspective to scheme, and crowd and new scene image are pieced together, generate new
Image be used for, concrete operations carry out in the steps below:
(1), the acquisition of dense population (prospect) image
Selecting current crowd to count personage's image data set currently in use is dense population image, including real scene image
And in image each personage head portrait centre coordinate location information, using front and back scape divide network model, what be will acquire is intensive
Crowd (prospect) image isolates crowd (prospect) image, and scene image using all kinds of keywords by being searched in a search engine
Rope scene image, or manually shoot on the spot, and artificial screening with suitably piece together to obtain the image not comprising personage.
(2), model obtains
The front and back scape segmentation neural network model for obtaining existing pre-training model and the search application based on vanishing Point Detection Method.
(3), crowd's (prospect) image obtains
Intensive character image is inputted in the scape segmentation network of front and back, crowd's foreground image is obtained, keeps initial markup information still
It is matched with the character positions of segmented image.
(4), alternate scenes figure determines
It is put into the image search system based on end point using intensive character image as query term, which passes through detection image
End point, searches out the picture candidate with similar end point in all scene images, these candidate images and original image will have
There is similar perspective view, enables the image ultimately generated more true.
(5), dense population image generates under new scene
For the crowd's foreground image and alternate scenes image of above-mentioned acquisition, wherein crowd's foreground image remains image and people
The matching of object location markup information carries out certain image and cuts scaling at random, and alternate scenes image collages, generates new close
Collection crowd's image, can be realized the express statistic of crowd's quantity under the scene.
Above only the present invention is further illustrated, and not to limit this patent, all is equivalence enforcement of the present invention,
It is intended to be limited solely by within the scope of the claims of this patent.
Claims (1)
1. a kind of dense population image generating method based on the segmentation of front and back scape, it is characterised in that front and back scape is used to divide network mould
Similar method is had an X-rayed in type and vanishing Point Detection Method search, and the crowd that cutting obtains is pieced together with new scene image, generates more
The dense population image of scape, realizes the express statistic of crowd's quantity under the scene, detailed process the following steps are included:
A step: the dense population image that will acquire utilizes front and back scape segmentation network model, isolates crowd and background, obtains people
Group's picture set, the dense population image are existing true intensive scene image；The background image is not comprising personage
Various scene pictures；
B step: using the similar image search method of end point, finds the scene for having similar perspective view to background image
As the alternate scenes picture pieced together；
Step c: crowd's picture and the alternate scenes picture in b step that a step obtains are pieced together, the intensive of more scenes is generated
Crowd's image, for crowd's quantity under the express statistic scene.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910480368.5A CN110276779A (en) | 2019-06-04 | 2019-06-04 | A kind of dense population image generating method based on the segmentation of front and back scape |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910480368.5A CN110276779A (en) | 2019-06-04 | 2019-06-04 | A kind of dense population image generating method based on the segmentation of front and back scape |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110276779A true CN110276779A (en) | 2019-09-24 |
Family
ID=67961888
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910480368.5A Pending CN110276779A (en) | 2019-06-04 | 2019-06-04 | A kind of dense population image generating method based on the segmentation of front and back scape |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110276779A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110852172A (en) * | 2019-10-15 | 2020-02-28 | 华东师范大学 | Method for expanding crowd counting data set based on Cycle Gan picture collage and enhancement |
| CN111860387A (en) * | 2020-07-27 | 2020-10-30 | 平安科技（深圳）有限公司 | Method and device for expanding data and computer equipment |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150230875A1 (en) * | 2014-02-17 | 2015-08-20 | Children's National Medical Center | Method and system for providing recommendation for optimal execution of surgical procedures |
| CN106327454A (en) * | 2015-07-03 | 2017-01-11 | 株式会社理光 | Method and device of synthesizing image |
| CN108460414A (en) * | 2018-02-27 | 2018-08-28 | 北京三快在线科技有限公司 | Generation method, device and the electronic equipment of training sample image |
| CN108564103A (en) * | 2018-01-09 | 2018-09-21 | 众安信息技术服务有限公司 | Data processing method and device |
| CN108804992A (en) * | 2017-05-08 | 2018-11-13 | 电子科技大学 | A kind of Demographics' method based on deep learning |
| CN108805201A (en) * | 2018-06-08 | 2018-11-13 | 湖南宸瀚信息科技有限责任公司 | Destination image data set creation method and its device |
-
2019
- 2019-06-04 CN CN201910480368.5A patent/CN110276779A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150230875A1 (en) * | 2014-02-17 | 2015-08-20 | Children's National Medical Center | Method and system for providing recommendation for optimal execution of surgical procedures |
| CN106327454A (en) * | 2015-07-03 | 2017-01-11 | 株式会社理光 | Method and device of synthesizing image |
| CN108804992A (en) * | 2017-05-08 | 2018-11-13 | 电子科技大学 | A kind of Demographics' method based on deep learning |
| CN108564103A (en) * | 2018-01-09 | 2018-09-21 | 众安信息技术服务有限公司 | Data processing method and device |
| CN108460414A (en) * | 2018-02-27 | 2018-08-28 | 北京三快在线科技有限公司 | Generation method, device and the electronic equipment of training sample image |
| CN108805201A (en) * | 2018-06-08 | 2018-11-13 | 湖南宸瀚信息科技有限责任公司 | Destination image data set creation method and its device |
Non-Patent Citations (2)
| Title |
|---|
| MUHAMMAD ARIF 等: "Counting of People in the Extremely Dense Crowd using Genetic Algorithm and Blobs Counting", 《IAES INTERNATIONAL JOURNAL OF ARTIFICIAL INTELLIGENCE》 * |
| YUKUN TIAN 等: "PaDNet: Pan-Density Crowd Counting", 《ARXIV》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110852172A (en) * | 2019-10-15 | 2020-02-28 | 华东师范大学 | Method for expanding crowd counting data set based on Cycle Gan picture collage and enhancement |
| CN110852172B (en) * | 2019-10-15 | 2020-09-22 | 华东师范大学 | Method for expanding crowd counting data set based on Cycle Gan picture collage and enhancement |
| CN111860387A (en) * | 2020-07-27 | 2020-10-30 | 平安科技（深圳）有限公司 | Method and device for expanding data and computer equipment |
| WO2021139340A1 (en) * | 2020-07-27 | 2021-07-15 | 平安科技（深圳）有限公司 | Data extension method and apparatus, and computer device |
| CN111860387B (en) * | 2020-07-27 | 2023-08-25 | 平安科技（深圳）有限公司 | Method, device and computer equipment for expanding data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ren et al. | Reciprocal transformations for unsupervised video object segmentation | |
| CN107885760B (en) | Knowledge graph representation learning method based on multiple semantics | |
| CN109684913A (en) | A kind of video human face mask method and system based on community discovery cluster | |
| CN109344702B (en) | Pedestrian detection method and device based on depth image and color image | |
| CN102142089B (en) | Semantic binary tree-based image annotation method | |
| CN110211108A (en) | A kind of novel abnormal cervical cells automatic identifying method based on Feulgen colouring method | |
| CN109213882A (en) | Picture sort method and terminal | |
| CN107222795A (en) | A kind of video abstraction generating method of multiple features fusion | |
| CN110348444A (en) | Wrong topic collection method, device and equipment based on deep learning | |
| CN109190588A (en) | A kind of method and device of population classification | |
| CN110276779A (en) | A kind of dense population image generating method based on the segmentation of front and back scape | |
| CN110136198A (en) | Image processing method and its device, equipment and storage medium | |
| Wu et al. | Movemine 2.0: Mining object relationships from movement data | |
| CN106033443B (en) | A kind of expanding query method and device in vehicle retrieval | |
| CN108268875A (en) | A kind of image meaning automatic marking method and device based on data smoothing | |
| CN104317946A (en) | Multi-key image-based image content retrieval method | |
| CN109272997A (en) | A kind of smart machine and its projection interactive method | |
| CN110532886A (en) | A kind of algorithm of target detection based on twin neural network | |
| CN105678244B (en) | A kind of near video search method based on improved edit-distance | |
| CN110309810A (en) | A kind of pedestrian's recognition methods again based on batch center similarity | |
| Liu et al. | Spatial pyramid block for oracle bone inscription detection | |
| CN107977592A (en) | A kind of image text detection method and system, user terminal and server | |
| CN109669967A (en) | A kind of space-time data association analysis method based on big data technology | |
| CN108121806A (en) | One kind is based on the matched image search method of local feature and system | |
| CN107357834A (en) | A kind of image search method of view-based access control model conspicuousness fusion |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190924 |