CN104268578A - Target recognition method implemented by combining small image comparison and fuzzy recognition - Google Patents

Target recognition method implemented by combining small image comparison and fuzzy recognition Download PDF

Info

Publication number
CN104268578A
CN104268578A CN201410541770.7A CN201410541770A CN104268578A CN 104268578 A CN104268578 A CN 104268578A CN 201410541770 A CN201410541770 A CN 201410541770A CN 104268578 A CN104268578 A CN 104268578A
Authority
CN
China
Prior art keywords
track
image
recognition
little image
people
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
Application number
CN201410541770.7A
Other languages
Chinese (zh)
Other versions
CN104268578B (en
Inventor
张鹏锐
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.)
SHENZHEN XIAOZHOU TECHNOLOGY Co Ltd
Original Assignee
SHENZHEN XIAOZHOU 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 SHENZHEN XIAOZHOU TECHNOLOGY Co Ltd filed Critical SHENZHEN XIAOZHOU TECHNOLOGY Co Ltd
Priority to CN201410541770.7A priority Critical patent/CN104268578B/en
Publication of CN104268578A publication Critical patent/CN104268578A/en
Application granted granted Critical
Publication of CN104268578B publication Critical patent/CN104268578B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00Scenes; Scene-specific elements
    • G06V20/50Context or environment of the image
    • G06V20/52Surveillance or monitoring of activities, e.g. for recognising suspicious objects
    • G06V20/53Recognition of crowd images, e.g. recognition of crowd congestion
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00Arrangements for image or video recognition or understanding
    • G06V10/94Hardware or software architectures specially adapted for image or video understanding

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a target recognition method implemented by combining small image comparison and fuzzy recognition. Defects of single-mode recognition are overcome, the accuracy rate of human head recognition can be increased by about 10% according to small-scale test results, and therefore the accuracy rate of customer flow is increased. The target recognition method can be applied to the field of customer flow statistics. By means of the target recognition method, the human head detection rates in different regions or at different angles under the same environment can be increased, and the human head detection rates under different environments can be increased. According to test results under an actual environment, the human head detection rate is increased by about 10% compared with that in the prior art. The human head detection rates are increased, more realistic trajectories can be constructed, and therefore the accuracy rate of customer flow statistics is increased. According to the test results under the actual environment, the accuracy rate of customer flow statistics is increased by about 5% compared with that in the prior art.

Description

A kind of little image ratio is to the target identification method combined with fuzzy diagnosis
Technical field
The present invention discloses a kind of target identification method, and particularly a kind of little image ratio is to the target identification method combined with fuzzy diagnosis, belongs to field of image recognition.
Background technology
Passenger flow statistics is a kind of data statistical approach based on image recognition, passenger flow statistics is relatively widely used in the solid shop/brick and mortar store such as business supermarket, shopping center, brand shop, 3C number, and for bus, long-distance big bus, subway gateway etc., can add up the volume of the flow of passengers.The Technical comparing of passenger flow statistics of the prior art is many, is mainly divided into following several: infrared emission technology, gravity sensing technology, WIFI detection technique, based on the moving object detecting of video with based on the mode identification technology of video.From the angle of accuracy rate, the mode identification technology that the highest is based on video.Pattern-recognition, by according to the video data collected, carries out pattern-recognition, obtain the coordinate of people, but consider differing greatly of actual scene, so, the Head recognition accuracy rate of being undertaken by algorithm for pattern recognition also cannot reach 100%, according to small-scale test result, can reach about 70%.
Summary of the invention
Accuracy rate for the above-mentioned Head recognition of the prior art mentioned is an important parameter of the accuracy rate affecting passenger flow statistics, a kind of method that invention provides little image recognition and pattern-recognition picture to combine, make up the deficiency of simple pattern-recognition, according to small-scale test result, the accuracy rate of Head recognition can promote about 10%, and then promotes the accuracy rate of passenger flow.
The present invention solves the technical scheme that its technical matters adopts: a kind of little image ratio is to the target identification method combined with fuzzy diagnosis, and the method comprises the steps:
Step S1: carry out image acquisition by video camera, the every two field picture collected can receive successively according to time sequencing;
Step S2: obtain image successively, and number of people detection is carried out to this two field picture, if the number of people do not detected, then jump to step S10, carry out the operation of step S10; If the number of people detected, then carry out the operation of step S3;
Step S3: the number of people coordinate preserved current frame number and detect, and preserve number of people coordinate intercept little image to little image buffer storage;
Step S4: according to the number of people coordinate position in this two field picture, be confirmed whether to be inserted in track and go, track can be inserted and have two conditions: first condition compares frame number, compare with the frame number of the node in track, if the interpolation of two frame numbers is less than or equal to 3, so just carry out the differentiation of second bar, the distance of the frame coordinate namely in the coordinate of number of people central point and track compares, if the distance of these two coordinates, be less than 40 pixels, then satisfy condition, these two conditions are all satisfied, just can be inserted in track; If do not met, then track can not be inserted;
Step S5: go if can be inserted in track, so will directly insert, otherwise, step S6 flow process will be proceeded to;
Step S6: judge to be inserted into the reason of going in track, if because the result of pattern-recognition, then proceed to step S8, by the flow process according to step S8, start the mode that little image ratio is right, if not because the result of pattern-recognition, then proceeds to step S7;
Step S7: can not be inserted into the reason of going in track, if the result of little image recognition, so will abandon this node;
Step S8: can not be inserted into the reason of going in track, if the result of pattern-recognition, is so directly inserted in track and goes;
Step S9: come back to S2.
The technical scheme that the present invention solves the employing of its technical matters further comprises:
In described step S1,25 two field pictures received recently are stored in buffer memory.
In described step S3,25 up-to-date two field pictures are preserved to the buffer memory of little image.
In described step S4 there is one or more in track, there is more than one target in a two field picture, needs the comparison of multi-to-multi.
Little image analysis mode is started in described step S8, confirm whether history image has the result of this time pattern-recognition in existence and this image to think similar image, if coupling finds similar image, step S4 will be jumped to, carry out the analysis of track, if do not find the image of coupling, so this result will be abandoned.
During described little image analysis mode, the threshold value setting of similarity is 12, and being namely less than threshold values 12, is exactly similar image.
The invention has the beneficial effects as follows: the present invention can be applicable to passenger flow statistics field, the number of people verification and measurement ratio of zones of different under equivalent environment or angle can be improved by the present invention, and improve the number of people verification and measurement ratio under varying environment.According to the test effect under actual environment, promote about about 10% than number of people verification and measurement ratio before.Improve number of people verification and measurement ratio, more real track can be formed, and then promote the accuracy rate of passenger flow statistics.According to the test effect under actual environment, promote about about 5% than passenger flow statistics accuracy rate before.
Below in conjunction with the drawings and specific embodiments, the present invention will be further described.
Accompanying drawing explanation
Fig. 1 is processing flow chart of the present invention.
Embodiment
The present embodiment is the preferred embodiment for the present invention, and other its principles all are identical with the present embodiment or approximate with basic structure, all within scope.
Please refer to accompanying drawing 1, the little image ratio of the present invention is to the target identification method combined with fuzzy diagnosis, and it specifically comprises the steps:
Step S1: carry out image acquisition by video camera, the every two field picture collected can receive successively according to time sequencing, so, in the present embodiment, first these images are numbered, from 0 frame, often accept a frame, its frame number can add 1, to 25 two field pictures received recently, be put in buffer memory (being called for short large image buffer storage, lower same) and go.
Step S2: algorithm for pattern recognition: obtain image successively from buffer memory, and number of people detection is carried out to this two field picture, number of people detection is the one of algorithm for pattern recognition, by the repeated segmentation to image, image in the little image split and pattern base is compared, extract the image that the similar number of people adds shoulder form, confirm that (common little image is rectangle for the size coordinate of little image, upper left and lower-right most point coordinate, center point coordinate or number of people coordinate are the center point coordinate of this rectangle), people's head inspecting method in the present embodiment can adopt people's head inspecting method of routine techniques.In the present embodiment, for recognition result You Liangge branch: one is if the number of people do not detected, then jump to step S9, carry out the operation of step S9; Another one branch detects the number of people, then carry out the branch operation of step S3.
Step S3: if the number of people detected, the number of people coordinate so then preserved current frame number and detect, and preserve little image to little image buffer storage (being called for short little image buffer storage, lower same), this buffer memory also will preserve 25 up-to-date two field pictures.
Step S4: according to the number of people coordinate position in this two field picture, be confirmed whether to be inserted in track and go, track in the present embodiment is multiple lists, a List Identification is a track, multiple node is comprised in a list, namely this node is the coordinate (i.e. number of people coordinate) of little picture centre, and the coordinate of two adjacent nodes must meet two conditions: frame number difference is less than or equal to 3; The distance of two node coordinates is less than 40 pixels.So judging that number of people coordinate can insert track and have two conditions: first condition compares frame number, compare with the frame number of the node in track, if the interpolation of two frame numbers is less than or equal to 3, so just carry out the differentiation of second bar, the distance of the number of people coordinate namely in number of people coordinate and track compares, if the distance of these two coordinates, be less than 40 pixels, then satisfy condition, these two conditions are all satisfied, just can be inserted in track.If do not met, this can not insert track, and track may exist many, equally, also can there is multiple target in a two field picture, so need the comparison of multi-to-multi.
Step S5: go if can be inserted in track, so goes being directly inserted in track, otherwise, step S6 flow process will be proceeded to.
Step S6: judge to be inserted into the reason of going in track, if because the result of pattern-recognition, then proceed to step S8, by the flow process according to step S8, start the mode that little image ratio is right, if not because the result of pattern-recognition, then proceeds to step S7.
Step S7: can not be inserted into the reason of going in existing track, if the result of little image recognition, so will abandon this node.
Step S8: can not be inserted into the reason of going in existing track, if the result of pattern-recognition, so will create new trajectory lists, as a new track.In addition, start little image analysis mode, little image analysis mode is a kind of algorithm contrasting two image similarities simply by aberration, and generally, an object is relatively stable at the aberration of same scene, so similarity is also high.In this branch, little image analysis mode mainly confirms that the image in large image buffer storage is compared with the little image of this pattern-recognition, be confirmed whether to there is similar image, in the present embodiment, the threshold value setting of similarity is 12, namely being less than 12 these threshold values, is exactly similar image, if coupling finds similar image, to step S4 be jumped to, carry out the analysis of track.If do not find the image of coupling, so this result will be abandoned.
S9: in time the number of people cannot being detected by pattern-recognition, will start little image analysis mode, compare to the image in current frame image and little image buffer storage, be confirmed whether to exist the similar image that similarity is less than 12.If existed, so continue the logic according to S4, carry out the analysis of track; If there is no, this result will be abandoned.
Overall description: so every two field picture all can circulate from S1 to S9, carries out the filtration of pattern-recognition and little image recognition two kinds of methods, promotes the verification and measurement ratio of target.In addition, in multiple lists of track, if find that certain list does not have new node city at 25 frames, can carry out analyzing and counting according to the Distance geometry angle of trajectory lists interior joint number, first node and tail node, after having counted, by this trajectory lists of deletion, certainly, this part is not the content that the present embodiment needs emphasis to describe.

Claims (6)

1. little image ratio is to the target identification method combined with fuzzy diagnosis, it is characterized in that: described method comprises the steps:
Step S1: carry out image acquisition by video camera, the every two field picture collected can receive successively according to time sequencing;
Step S2: obtain image successively, and number of people detection is carried out to this two field picture, if the number of people do not detected, then jump to step S10, carry out the operation of step S10; If the number of people detected, then carry out the operation of step S3;
Step S3: the number of people coordinate preserved current frame number and detect, and preserve number of people coordinate intercept little image to little image buffer storage;
Step S4: according to the number of people coordinate position in this two field picture, be confirmed whether to be inserted in track and go, track can be inserted and have two conditions: first condition compares frame number, compare with the frame number of the node in track, if the interpolation of two frame numbers is less than or equal to 3, so just carry out the differentiation of second bar, the distance of the frame coordinate namely in the coordinate of number of people central point and track compares, if the distance of these two coordinates, be less than 40 pixels, then satisfy condition, these two conditions are all satisfied, just can be inserted in track; If do not met, then track can not be inserted;
Step S5: go if can be inserted in track, so will directly insert, otherwise, step S6 flow process will be proceeded to;
Step S6: judge to be inserted into the reason of going in track, if because the result of pattern-recognition, then proceed to step S8, by the flow process according to step S8, start the mode that little image ratio is right, if not because the result of pattern-recognition, then proceeds to step S7;
Step S7: can not be inserted into the reason of going in track, if the result of little image recognition, so will abandon this node;
Step S8: can not be inserted into the reason of going in track, if the result of pattern-recognition, is so directly inserted in track and goes;
Step S9: come back to step S2.
2. little image ratio according to claim 1 is to the target identification method combined with fuzzy diagnosis, it is characterized in that: be stored in buffer memory 25 two field pictures received recently in described step S1.
3. little image ratio according to claim 1 is to the target identification method combined with fuzzy diagnosis, it is characterized in that: preserve 25 up-to-date two field pictures to the buffer memory of little image in described step S3.
4. little image ratio according to claim 1 is to the target identification method combined with fuzzy diagnosis, it is characterized in that: in described step S4, track exists one or more, there is more than one target in a two field picture, need the comparison of multi-to-multi.
5. little image ratio according to claim 1 is to the target identification method combined with fuzzy diagnosis, it is characterized in that: in described step S8, start little image analysis mode, confirm whether history image has the result of this time pattern-recognition in existence and this image to think similar image, if coupling finds similar image, to step S4 be jumped to, carry out the analysis of track, if, do not find the image of coupling, so will abandon this result.
6. little image ratio according to claim 5 is to the target identification method combined with fuzzy diagnosis, it is characterized in that: during described little image analysis mode, the threshold value setting of similarity is 12, and being namely less than threshold values 12, is exactly similar image.
CN201410541770.7A 2014-10-15 2014-10-15 The target identification method that a kind of small image is compared and fuzzy diagnosis is combined Active CN104268578B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201410541770.7A CN104268578B (en) 2014-10-15 2014-10-15 The target identification method that a kind of small image is compared and fuzzy diagnosis is combined

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201410541770.7A CN104268578B (en) 2014-10-15 2014-10-15 The target identification method that a kind of small image is compared and fuzzy diagnosis is combined

Publications (2)

Publication Number Publication Date
CN104268578A true CN104268578A (en) 2015-01-07
CN104268578B CN104268578B (en) 2017-06-09

Family

ID=52160098

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201410541770.7A Active CN104268578B (en) 2014-10-15 2014-10-15 The target identification method that a kind of small image is compared and fuzzy diagnosis is combined

Country Status (1)

Country Link
CN (1) CN104268578B (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109448026A (en) * 2018-11-16 2019-03-08 南京甄视智能科技有限公司 Passenger flow statistical method and system based on head and shoulder detection
CN115984948A (en) * 2023-03-20 2023-04-18 广东广新信息产业股份有限公司 Face recognition method applied to temperature sensing and electronic equipment
CN116386106A (en) * 2023-03-16 2023-07-04 宁波星巡智能科技有限公司 Intelligent infant head recognition method, device and equipment during sleep-accompanying infant

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110164816A1 (en) * 2010-01-05 2011-07-07 Canon Kabushiki Kaisha Image processing apparatus, image processing method, and storage medium
CN104077804A (en) * 2014-06-09 2014-10-01 广州嘉崎智能科技有限公司 Method for constructing three-dimensional human face model based on multi-frame video image

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110164816A1 (en) * 2010-01-05 2011-07-07 Canon Kabushiki Kaisha Image processing apparatus, image processing method, and storage medium
CN104077804A (en) * 2014-06-09 2014-10-01 广州嘉崎智能科技有限公司 Method for constructing three-dimensional human face model based on multi-frame video image

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
陈艳艳等: "基于机器视觉的地铁站客流自动检测方法", 《中国期刊全文数据库 公路交通科技》 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109448026A (en) * 2018-11-16 2019-03-08 南京甄视智能科技有限公司 Passenger flow statistical method and system based on head and shoulder detection
CN116386106A (en) * 2023-03-16 2023-07-04 宁波星巡智能科技有限公司 Intelligent infant head recognition method, device and equipment during sleep-accompanying infant
CN115984948A (en) * 2023-03-20 2023-04-18 广东广新信息产业股份有限公司 Face recognition method applied to temperature sensing and electronic equipment

Also Published As

Publication number Publication date
CN104268578B (en) 2017-06-09

Similar Documents

Publication Publication Date Title
CN107527009B (en) Remnant detection method based on YOLO target detection
CN105938622B (en) Method and apparatus for detecting object in moving image
CN107798272B (en) Rapid multi-target detection and tracking system
CN107657244B (en) Human body falling behavior detection system based on multiple cameras and detection method thereof
CN106991370B (en) Pedestrian retrieval method based on color and depth
JP6095817B1 (en) Object detection device
CN105940430A (en) Person counting method and device for same
CN105893957B (en) View-based access control model lake surface ship detection recognition and tracking method
CN109255386B (en) Road pedestrian rapid detection method based on millimeter wave radar and vision fusion
CN109712171B (en) Target tracking system and target tracking method based on correlation filter
CN102610104B (en) Onboard front vehicle detection method
CN104268595B (en) general object detection method and system
CN104268578A (en) Target recognition method implemented by combining small image comparison and fuzzy recognition
CN106446832B (en) Video-based pedestrian real-time detection method
KR101501531B1 (en) Stereo Vision-based Pedestrian Detection System and the method of
CN108664853B (en) Face detection method and device
CN113449629B (en) Lane line false and true identification device, method, equipment and medium based on driving video
CN106991684B (en) Foreground extracting method and device
Chen et al. Detection and segmentation of occluded vehicles based on skeleton features
Luo et al. A real-time people counting approach in indoor environment
CN103955925A (en) Improved probability Hough transform curve detection method based on partitioning minimum samples fixing
CN109977740B (en) Depth map-based hand tracking method
CN108009504B (en) Moving sphere identification method and device and storage medium
CN105930813A (en) Method for detecting line text under any natural scene
CN112308023B (en) Method, device and equipment for identifying pedestrian movement

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