CN113627365A - Group movement identification and time sequence analysis method - Google Patents
Group movement identification and time sequence analysis method Download PDFInfo
- Publication number
- CN113627365A CN113627365A CN202110935928.9A CN202110935928A CN113627365A CN 113627365 A CN113627365 A CN 113627365A CN 202110935928 A CN202110935928 A CN 202110935928A CN 113627365 A CN113627365 A CN 113627365A
- Authority
- CN
- China
- Prior art keywords
- auction
- action
- time sequence
- identification
- human body
- 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
- 230000033001 locomotion Effects 0.000 title claims abstract description 35
- 238000000034 method Methods 0.000 title claims abstract description 35
- 238000012300 Sequence Analysis Methods 0.000 title claims abstract description 15
- 238000012545 processing Methods 0.000 claims description 9
- 238000004873 anchoring Methods 0.000 claims description 8
- 238000012797 qualification Methods 0.000 claims description 7
- 230000000007 visual effect Effects 0.000 claims description 4
- 230000000903 blocking effect Effects 0.000 claims description 2
- 238000005516 engineering process Methods 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 3
- 238000010606 normalization Methods 0.000 description 3
- 230000002708 enhancing effect Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/08—Auctions
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Finance (AREA)
- Data Mining & Analysis (AREA)
- Accounting & Taxation (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Entrepreneurship & Innovation (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a group motion recognition and time sequence analysis method, which is a method for rapidly, efficiently and accurately recognizing group motions in an auction scene and determining the priority of an auctioneer through analyzing the motion time sequence, wherein human body joint points are replaced by an anchored human body center point and an anchored auction board center point, and continuous track prediction is replaced by discrete key frame prediction; aiming at the characteristics that the auction scene has more people and more shelters but the change of the position of the human body is small, after the prior art and the existing requirements are comprehensively considered, a group movement identification and time sequence analysis method is designed. The rapid auction action recognition method provided by the invention has both accuracy and timeliness, and can reduce the requirements on hardware equipment.
Description
Technical Field
The invention relates to the field of computer vision, in particular to a group motion identification and time sequence analysis method.
Technical Field
The human body action recognition method mainly comprises two types, the first type is recognition of time sequence action constructed by a series of continuous image frames through pipeline linkage based on image processing, the accuracy rate of the method is high, but the calculated amount is large, and the operation speed is low. The second category is skeleton-based, where the skeleton is first determined, how to identify the actions according to the spatio-temporal characteristics of the skeleton, and the timing is determined. The method can well cope with the shielding situation and has good generalization, but the method has lower accuracy and higher training difficulty when facing the action recognition of population.
Disclosure of Invention
The purpose of the invention is as follows: aiming at the existing requirements of a specific auction scene, a group movement recognition and time sequence analysis method is designed by utilizing the existing technical means, and convenience is provided for managers to lock auction willingness persons in advance and intelligently determine the priority of the auction persons. The novel auction mode taking computer vision as a support and a public large screen as a carrier is beneficial to enhancing the openness and transparency of the auction, relieving the working pressure by science and technology and benefiting people by science and technology.
The method integrates the advantages of motion recognition based on image processing and motion recognition based on a skeleton, replaces human body joint points with anchored human body center points and anchored auction board center points, and replaces continuous track prediction with discrete key frame prediction. The method has the characteristics of small calculated amount, high timeliness and high accuracy; the influence on the environment is small under the condition of multiple persons and sheltering. And the time sequence information of the action is also concerned, and the priority of the auction participants is determined through the optimization judgment of the start-stop frame, so that the model has the value of engineering application.
The technical scheme is that the group motion identification and time sequence analysis method comprises the following steps:
step S1: positioning the human body and the auction plate, and determining the central points of the anchoring frames of the human body and the auction plate to form a connecting line;
step S2: selecting a proper base line, taking an included angle parameter formed by a central connecting line of the two base lines as a main action judgment basis, and taking a vertical height parameter as an auxiliary basis for action identification;
step S3: determining a discrete key frame as a prediction basis;
step S4: determining the auction qualification priority of the auctioneers according to the change rate of the angular speed in the time sequence;
step S5: continuously collecting action image frames, and establishing an action recognition model according to the change characteristics of action discrimination parameters corresponding to body actions;
furthermore, in step S1, since there are many personnel in the auction site and there is a high possibility that personnel are blocked, but the seat intervals in the auction site are relatively regular, when positioning the personnel, the positioning of the human body is cross-verified according to the preset space intervals in addition to the positioning of the human body by recognition, and if the body part of the human body is recognized, it is determined that there is personnel at that place, so that the influence of the blocking phenomenon on the identification of the auction is reduced.
Further, in step S2, the central connection line needs to form an included angle with the baseline, and then is used as a main judgment basis for the auction action; selecting as a baseline a line parallel to the ground and parallel to the camera's visual plane; and meanwhile, the normalized vertical height parameter is taken as an auxiliary basis for motion recognition.
Furthermore, in step S3, because of the rapidity of the auction action, a higher requirement is placed on the processing performance of the computer, and under the condition that the processing capability of the computer is fixed, in order to improve the real-time performance of the model, a prediction module is added, and because only whether the action belongs to the auction action needs to be determined, the method is different from the traditional continuous trajectory prediction, and a discrete key frame prediction method is designed.
Further, the discrete key frame prediction method comprises: the discrete representative key frames are selected as the basis of motion prediction, the sin value of the arm angle of a quasi-auction player in a general state is a negative number, the sin value is greater than 1/2 when an obvious auction intention exists, and the sin value is about 1 when the auction action stops, so that after the auction intention of the auction player is judged, the model does not need to continuously identify the motion of an image frame any more, but compares the discrete key frames with the key frames according to a certain time sequence at certain intervals to reduce the calculated amount and improve the timeliness of the model.
Further, several representative key frames are 30 °, 40 °, 50 °, 60 °, 70 °, 80 °, respectively.
Further, in step S4, the purpose of the identification of the auction actions is not only to identify the actions, but also to determine the order of the participants for engineering use, and therefore, the priority of the qualification of the participants for auction is determined according to the rate of change of the angular velocity in the same time sequence set.
Further, in step S5, motion image frames are continuously collected, and a motion recognition model is established based on the variation characteristics of the motion discrimination parameters corresponding to the body motion.
Compared with the prior art, the group movement identification and time sequence analysis method has the following beneficial effects: the rapid auction action identification method has both accuracy and timeliness, can reduce the requirements on hardware equipment, and provides convenience for managers to lock auction willingness persons in advance and intelligently determine the priority of the auction persons. The novel auction mode taking computer vision as a support and a public large screen as a carrier is beneficial to enhancing the openness and transparency of the auction, relieving the working pressure by science and technology and benefiting people by science and technology.
Drawings
FIG. 1 is a flow chart of a method for identifying a group action;
FIG. 2 is a schematic illustration of a baseline selection;
FIG. 3 is a schematic view of height normalization;
FIG. 4 is a key frame diagram of an auction action.
Detailed Description
Based on visual image processing, group motions in an auction scene are quickly, efficiently and accurately identified, and the priorities of auction participants are determined by analyzing motion time sequences, so that a group motion identification and time sequence analysis method is designed after comprehensive consideration is given to the prior art and the existing requirements aiming at the characteristics that the auction scene has more unique people, more shelters and small change of human body positions, and the method comprises the following steps: anchoring the positions of the human body and the auction plate, and determining the central points of the human body and the auction plate to form a connecting line; selecting a proper base line, and taking an included angle parameter formed by a central connecting line of the two base lines as a main action identification parameter and a vertical height parameter as an auxiliary action identification parameter; selecting a discrete key frame as a motion prediction basis; judging the starting and ending frames of the action, and determining the auction qualification priority of the auctioneer according to the time sequence relation of the action; and establishing a motion recognition model according to the change characteristics of the motion discrimination parameters corresponding to the body motion. The rapid auction action recognition method provided by the invention has both accuracy and timeliness, and can reduce the requirements on hardware equipment.
As shown in fig. 1-4, a group motion recognition and time sequence analysis method includes:
step S1: positioning the human body and the auction plate, and determining the central points of the anchoring frames of the human body and the auction plate to form a connecting line, wherein the central points are determined through the anchoring frames obtained by positioning the human body and positioning the auction plate; because the number of personnel in the auction site is large, the possibility that the personnel are shielded exists, but the seat intervals in the auction site are relatively regular, when the personnel are positioned, the positioning of the human body is identified, the cross verification is also carried out on the positioning of the human body according to the preset space intervals, if the body part of the human body is identified, the personnel are judged to exist at the side, and the influence of the shielding phenomenon on the identification of the auction is reduced;
step S2: selecting a proper base line, taking an included angle parameter formed by a central connecting line of the two base lines as a main action judgment basis, and taking a vertical height parameter as an auxiliary basis for action identification; the central connecting line can be used as a main judgment basis for the auction action after forming an included angle with the base line, so that a line parallel to the ground and the visual surface of the camera is selected as the base line, as shown in fig. 2, and meanwhile, the normalized vertical height parameter H is taken as an auxiliary basis for action recognition, and normalization is relative normalization; h' = H1/H2, wherein H1 is the vertical distance from the anchoring center of the auction card to the center of the human body anchoring point, and H2 is the vertical distance from the center of the human body anchoring point to the center of the bottom end of the human body anchoring point, as shown in fig. 3;
step S3: and selecting discrete key frames as a prediction basis, wherein high requirements are put forward on the processing performance of the computer due to the rapidity of the auction action. Under the condition of certain computer processing capacity, a prediction module is added to improve the real-time performance of the model. The invention only needs to judge whether the action is the auction action, so that the invention is different from the traditional continuous track prediction and designs a discrete prediction method. A few discrete representative key frames are selected as the basis for motion prediction, as shown in FIG. 4, the sin value of the arm angle of the quasi-auction user under normal conditions is negative, the sin value is greater than 1/2 when there is an obvious auction intention, and the sin value is about 1 when the auction stops. Therefore, 30 degrees, 40 degrees, 50 degrees, 60 degrees, 70 degrees and 80 degrees are selected as key frames, when the included angle is 30 degrees, the auction intention of an auction player is judged, then the model does not continuously conduct action recognition on image frames any more, the image frames are compared with the key frames discretely according to a certain interval time sequence, the sin value of the angle is used as the confidence coefficient of the action recognition, and the confidence coefficient is higher if the sin value is high. Therefore, the calculation amount can be reduced, and the timeliness of the model is improved:
step S4: determining the auction qualification priority of the auctioneers according to the change rate of the angular speed in the time sequence; the purpose of the identification of the auction actions is not only identification of the actions, but also determination of the sequence of the auctioneers for engineering application. Therefore, the auction qualification priority of the auctioneers is determined according to the change rate of the angular speed in the same set time sequence. And selecting the image frame at 30 degrees as an initial frame and the image frame at 90 degrees as a termination frame, and determining the priority of the auction participants according to the number of the image frames passing through the angular change. The smaller the number of frames, the shorter the description time, and the higher the priority. Because the sin value can always keep a higher level when the action is stopped, and the judgment of the time sequence is influenced, an angle change slope is introduced, when a person keeps the auction ending action, the angle change is about 0, or the person oscillates between positive and negative, the person is judged to be an ending frame;
step S5: and continuously collecting action image frames, and establishing an action recognition model according to the change characteristics of the action discrimination parameters corresponding to the body action.
Claims (7)
1. A group motion identification and time sequence analysis method is characterized by comprising the following steps:
step S1: positioning the human body and the auction plate, and determining the central points of the anchoring frames of the human body and the auction plate to form a connecting line;
step S2: selecting a proper base line, taking an included angle parameter formed by a central connecting line of the two base lines as a main action judgment basis, and taking a vertical height parameter as an auxiliary basis for action identification;
step S3: determining a discrete key frame as a prediction basis;
step S4: determining the auction qualification priority of the auctioneers according to the change rate of the angular speed in the time sequence;
step S5: and continuously collecting action image frames, and establishing an action recognition model according to the change characteristics of the action discrimination parameters corresponding to the body action.
2. The group movement identification and time sequence analysis method according to claim 1, wherein in step S1, since there are many personnel in the auction site and there is a high possibility that the personnel are blocked, but the seat intervals in the auction site are relatively regular, when the personnel are positioned, the positioning of the human body is cross-verified according to the preset space intervals in addition to the identification of the human body, and if the body part of the human body is identified, the existence of the personnel at that side is determined, so as to reduce the influence of the blocking phenomenon on the identification of the auction.
3. The group movement identification and time sequence analysis method of claim 1, wherein in step S2, the central connection line needs to form an included angle with the baseline, and then is used as the main judgment basis for the auction action; selecting as a baseline a line parallel to the ground and parallel to the camera's visual plane; and meanwhile, the normalized vertical height parameter is taken as an auxiliary basis for motion recognition.
4. The group movement identification and time sequence analysis method according to claim 1, wherein in step S3, due to the rapidity of the auction action, a higher requirement is made on the processing performance of the computer, under the condition of a certain processing capability of the computer, in order to improve the real-time performance of the model, a prediction module is added, and since it is only necessary to judge whether the action belongs to the auction action, a discrete key frame prediction method is designed, which is different from the conventional continuous trajectory prediction.
5. The method of claim 4, wherein the discrete key frame prediction method comprises: the discrete representative key frames are selected as the basis of motion prediction, the sin value of the arm angle of a quasi-auction player in a general state is a negative number, the sin value is greater than 1/2 when an obvious auction intention exists, and the sin value is about 1 when the auction action stops, so that after the auction intention of the auction player is judged, the model does not need to continuously identify the motion of an image frame any more, but compares the discrete key frames with the key frames according to a certain time sequence at certain intervals to reduce the calculated amount and improve the timeliness of the model.
6. The method of claim 5, wherein the number of representative keyframes is 30 °, 40 °, 50 °, 60 °, 70 °, 80 °.
7. The group sport identification and time sequence analysis method as claimed in claim 1, wherein in step S4, the purpose of identification of the auction actions is not only identification of actions, but also determining the priority of the auctioneers for engineering application, so that the auction qualification priority of the auctioneers is determined according to the rate of change of angular velocity in the same set time sequence.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110935928.9A CN113627365A (en) | 2021-08-16 | 2021-08-16 | Group movement identification and time sequence analysis method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110935928.9A CN113627365A (en) | 2021-08-16 | 2021-08-16 | Group movement identification and time sequence analysis method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113627365A true CN113627365A (en) | 2021-11-09 |
Family
ID=78385571
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110935928.9A Pending CN113627365A (en) | 2021-08-16 | 2021-08-16 | Group movement identification and time sequence analysis method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113627365A (en) |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107886544A (en) * | 2016-09-30 | 2018-04-06 | 法乐第(北京)网络科技有限公司 | IMAQ control method and device for vehicle calibration |
CN112109664A (en) * | 2019-06-20 | 2020-12-22 | 上海汽车集团股份有限公司 | Control method and system |
CN112203115A (en) * | 2020-10-10 | 2021-01-08 | 腾讯科技(深圳)有限公司 | Video identification method and related device |
US20210110146A1 (en) * | 2019-10-15 | 2021-04-15 | Fujitsu Limited | Action recognition method and apparatus and electronic equipment |
CN112686153A (en) * | 2020-12-30 | 2021-04-20 | 西安邮电大学 | Three-dimensional skeleton key frame selection method for human behavior recognition |
CN113065505A (en) * | 2021-04-15 | 2021-07-02 | 中国标准化研究院 | Body action rapid identification method and system |
-
2021
- 2021-08-16 CN CN202110935928.9A patent/CN113627365A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107886544A (en) * | 2016-09-30 | 2018-04-06 | 法乐第(北京)网络科技有限公司 | IMAQ control method and device for vehicle calibration |
CN112109664A (en) * | 2019-06-20 | 2020-12-22 | 上海汽车集团股份有限公司 | Control method and system |
US20210110146A1 (en) * | 2019-10-15 | 2021-04-15 | Fujitsu Limited | Action recognition method and apparatus and electronic equipment |
CN112203115A (en) * | 2020-10-10 | 2021-01-08 | 腾讯科技(深圳)有限公司 | Video identification method and related device |
CN112686153A (en) * | 2020-12-30 | 2021-04-20 | 西安邮电大学 | Three-dimensional skeleton key frame selection method for human behavior recognition |
CN113065505A (en) * | 2021-04-15 | 2021-07-02 | 中国标准化研究院 | Body action rapid identification method and system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109934176B (en) | Pedestrian recognition system, recognition method, and computer-readable storage medium | |
CN110852219B (en) | Multi-pedestrian cross-camera online tracking system | |
CN101577812B (en) | Method and system for post monitoring | |
CN102831439B (en) | Gesture tracking method and system | |
CN100390811C (en) | Method for tracking multiple human faces from video in real time | |
CN106845385A (en) | The method and apparatus of video frequency object tracking | |
CN104361327A (en) | Pedestrian detection method and system | |
CN105574506A (en) | Intelligent face tracking system and method based on depth learning and large-scale clustering | |
CN105426827A (en) | Living body verification method, device and system | |
JP2007334623A (en) | Face authentication device, face authentication method, and access control device | |
CN101281646A (en) | Method for real-time detection of driver fatigue based on vision | |
CN111832400A (en) | Mask wearing condition monitoring system and method based on probabilistic neural network | |
CN114100103B (en) | Rope skipping counting detection system and method based on key point identification | |
CN110750152A (en) | Human-computer interaction method and system based on lip action | |
CN114140745A (en) | Method, system, device and medium for detecting personnel attributes of construction site | |
CN112700568B (en) | Identity authentication method, equipment and computer readable storage medium | |
CN110188718A (en) | It is a kind of based on key frame and joint sparse indicate without constraint face identification method | |
CN113627365A (en) | Group movement identification and time sequence analysis method | |
CN115841497B (en) | Boundary detection method and escalator area intrusion detection method and system | |
CN101393602B (en) | Be applied to the eye identifying method for tracing of personnel control | |
CN116703968A (en) | Visual tracking method, device, system, equipment and medium for target object | |
CN113554682B (en) | Target tracking-based safety helmet detection method | |
CN206610308U (en) | A kind of data processing equipment that face recognition is realized by eyes orientation direction | |
CN115731563A (en) | Method for identifying falling of remote monitoring personnel | |
CN110060756A (en) | A kind of more people's online management systems and method of interval training |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20211109 |
|
RJ01 | Rejection of invention patent application after publication |