CN106446837B - A kind of detection method of waving based on motion history image - Google Patents

A kind of detection method of waving based on motion history image Download PDF

Info

Publication number
CN106446837B
CN106446837B CN201610859376.7A CN201610859376A CN106446837B CN 106446837 B CN106446837 B CN 106446837B CN 201610859376 A CN201610859376 A CN 201610859376A CN 106446837 B CN106446837 B CN 106446837B
Authority
CN
China
Prior art keywords
waving
pedestrian
image
detection
frame
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201610859376.7A
Other languages
Chinese (zh)
Other versions
CN106446837A (en
Inventor
罗文峰
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.)
Hunan Youxiang Technology Co Ltd
Original Assignee
Hunan Youxiang 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 Hunan Youxiang Technology Co Ltd filed Critical Hunan Youxiang Technology Co Ltd
Priority to CN201610859376.7A priority Critical patent/CN106446837B/en
Publication of CN106446837A publication Critical patent/CN106446837A/en
Application granted granted Critical
Publication of CN106446837B publication Critical patent/CN106446837B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/20—Movements or behaviour, e.g. gesture recognition
    • G06V40/28—Recognition of hand or arm movements, e.g. recognition of deaf sign language
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/107—Static hand or arm
    • 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
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
    • G06V10/507—Summing image-intensity values; Histogram projection analysis

Landscapes

  • Engineering & Computer Science (AREA)
  • Human Computer Interaction (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Theoretical Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Health & Medical Sciences (AREA)
  • Psychiatry (AREA)
  • Social Psychology (AREA)
  • Image Analysis (AREA)

Abstract

The detection method of waving based on motion history image that the invention proposes a kind of, determine the region for detection of waving indirectly by pedestrian detection first, then the movement principal direction that the region is calculated by motion history image, finally judges the generation in sequence image whether there is or not waving motion.Method complexity proposed by the present invention is small, and accuracy in detection is high, is suitable for unmanned plane target with the goal verification stage before clapping.

Description

A kind of detection method of waving based on motion history image
Technical field
The present invention relates to image procossings, computer vision processing technology field, in particular to a kind of to be based on motion history figure The detection method of waving of picture.
Background technique
With the rapid development of automatic technology, the function of unmanned plane is stronger and stronger, and auto-tracking shooting target becomes One important component of unmanned plane application.People or object of the unmanned plane with needing selected tracking before clapping.In traditional target In selection, the method selected target of manual picture frame is generallyd use.This Method And Principle is simple, is easier to realize, but also has Apparent defect: 1, manual picture frame is inaccurate, the qualification used dependent on operator equipment and software；2, picture frame is past It is realized toward in handheld device such as mobile phone, is unable to automatic identification in the case where being detached from mobile phone.
In order to realize the Object selection for automatically tracking pedestrian, a new thinking is that the object of tracking is allowed to carry out certain behavior For example wave, it is then detected by behavior (such as waving) to determine to track target.
Behavioral value is an important topic of computer application, human-computer interaction, in terms of have extensively Application prospect.Its development course, mainly using hardware device as front end, obtains initial motion information in early days, is such as used for hand The data glove of movement detection systems.Often price is more expensive for hardware device needed for such methods, and need wear so that The comfort level of user is not high.
With the development of computer vision technique, software-based hand exercise detection technique also starts to develop.Hand at present Portion's movement detection systems process flow is broadly divided into two parts, and first part is that information needed is extracted from video sequence frame, this In information needed refer mainly to relative position of the moving object in video frame, method mainly has a colour of skin division at present, foreground partition, Solid division etc..And second part is such as implicit Ma Er Kraft chain, wavelet transformation, BP neural network by computation model Method, to handle previously obtained characteristics of image.
Hand exercise detection at present is mainly the combination of many algorithms, generally requires hand-characteristic extraction algorithm and movement is examined Method of determining and calculating combines.For identifying more demanding application, need to extract more features and complicated detection algorithm.And unmanned plane Target detection algorithm of waving with bat needs to run on mobile phone, can not provide higher computing capability, therefore existing method Be not suitable for unmanned plane to wave to detect with the target of bat.
Summary of the invention
For the deficiency of existing detection method of waving, the present invention proposes a kind of side of detection of waving based on motion history image Method.
The technical solution adopted by the present invention is that:
A kind of detection method of waving based on motion history image, comprising the following steps:
S1 trains a pedestrian detector；
Positive template is acquired, and is normalized；Negative sample is collected later；Then choice direction histogram of gradients (HOG) feature extracting method carries out feature extraction to positive negative sample, is trained using svm；
S2 carries out pedestrian detection when unmanned plane carries out target with clapping, to the video image taken, obtains according to classifier Divide height to be ranked up pedestrian candidate frame, chooses pair of the pedestrian candidate frame of wherein highest scoring as detection of waving for the first time As；
Pedestrian detection is carried out by the way of sliding window；
S3 carries out detection zone determination of waving to the highest pedestrian candidate frame of present score, then logical in detection zone of waving Motion history image is crossed to carry out waving to detect, which is characterized in that
Determine that one is waved to detect as the window of pedestrian detection according to the highest pedestrian candidate frame of present score first Window, detection window size of waving is 36*36, positioned at the upper left side of pedestrian detection window；It waves detection window and pedestrian detection The left vertex of window differs 12 pixels in x, y-axis respectively；
The video image for remembering unmanned plane shooting is { Pn(x, y) | n=1,2 ... N }, to each width image in video image, All only retain the pixel value waved in detection window region, even all pixels value for detection window exterior domain of waving is set to 0；
Since n-th (n >=2) frame image, the motion detection of n-th (n >=2) frame image is carried out using 3 frame differences, is calculated public Formula is as follows:
Dn(x, y)=Pn-1(x,y)-2Pn(x,y)+Pn+1(x,y) (1)
Wherein Pn(x,y)、Pn-1(x,y)、Pn+1(x, y) respectively indicates n-th frame image, the (n-1)th frame image, the (n+1)th frame figure Picture；
Then use classical big law by Dn(x, y) binaryzation obtains two-value motion information An(x,y)；
Next it carries out waving to detect by motion history image waving detection zone, the method is as follows:
A. motion history image H is definedn(x, y), wherein the value of each pixel is the run duration letter of the pixel Breath, is defined as follows:
B. on motion history image, with any pixel point (x0,y0) centered on, one shares 9 in the field that size is 3 × 3 A pixel just thinks pixel (x when 9 pixels are not zero0,y0) it is not boundary pixel；Otherwise it is assumed that the pixel Point (x0,y0) it is boundary pixel；
When calculating gradient direction, if pixel is boundary pixel, its gradient direction is directly set to 0；If pixel It is not boundary pixel, then its gradient direction namely angle matrix θ is calculated using sobel operatorn(x, y), as follows:
Wherein It is volume Product operation；
C. the range for the angle matrix being calculated is 0 to 359, use size for 36 histogram, successively to angle square Battle array θn(x, y) each pixel is traversed, statistic histogram.Then using angle corresponding to the maximum value of histogram as the figure The movement principal direction ω of picturen,
D. image { the P come is transmitted through from video endn(x, y) | n=1,2 ... N in the second frame image start, according to above-mentioned Method successively calculates the movement principal direction { ω of each imagen| n=2 ..., N }, then calculate a direction index value:
Wherein movement principal direction is judged as that the right side is waved between 46 degree and 134 degree；Principal direction is moved in 226 degree and 314 degree Between, it is judged as that a left side is waved；
When non-zero value reaches N/2 in current N width image, then determine that there are waving motions, otherwise, it is determined that not wave Hands movement；
When detecting waving motion, overall flow terminates S4；If the pedestrian candidate frame of highest scoring does not detect It waves behavior, then the pedestrian candidate frame of score second in the classifier score height ranking results of pedestrian detection in S2 is waved Hand detection, detection method of waving are identical with the detection method of waving for test object of waving for the first time；It is regular according to this, until detecting There is behavior of waving in some pedestrian candidate frame or all pedestrian candidate frames have carried out waving to detect not waving still Until behavior.
In S1 of the invention, the training method of pedestrian detector is: by the camera on unmanned plane to different pedestrians into Row shooting, for the pedestrian image of acquisition no less than 500 width as positive template, the pedestrian's number for participating in acquisition is no less than 100 people altogether. Then from network or other various databases collect it is various do not include pedestrian 200 width more than image as negative sample.It will Collected pedestrian image normalizes to the image that size is 108*36, selects classical histograms of oriented gradients (HOG) feature Extracting method carries out feature extraction to positive negative sample, is trained using svm, obtains a pedestrian detector.(illustrate herein: HOG+SVM is that French researcher Dalal is proposed on 2005 CVPR, is used for pedestrian detection, and the present invention uses this completely A method.)
The method of S2 of the present invention are as follows: when unmanned plane carries out target with clapping, pedestrian is carried out to the video image of shooting first Detection, pedestrian detection process use the size of sliding window for 108*36, extract the HOG feature of image in frame to be checked, pass through row People's detector is classified, obtain one whether be pedestrian score, if the score be greater than 0.7, with this frame to be detected be wait Select frame.When there are multiple candidate frames, pedestrian candidate frame is ranked up according to classifier score height, chooses wherein that score is most Object of the high pedestrian candidate frame as detection of waving for the first time.
N generally takes 30.
When non-zero value reaches 15 in current 30 width image, present invention determine that there are waving motions, otherwise, it is determined that for not There is waving motion.
The detection method of waving based on motion history image that the invention proposes a kind of, it is indirectly true by pedestrian detection first Surely it waves the region detected, the movement principal direction in the region then to be calculated by motion history image, finally judges sequence image In whether there is or not the generations of waving motion.Method complexity proposed by the present invention is small, and accuracy in detection is high, is suitable for unmanned plane target With the goal verification stage before clapping.
Detailed description of the invention
Fig. 1 is flow diagram of the invention；
Fig. 2 is the detection method schematic diagram of waving based on motion history image；
Fig. 3 is the schematic diagram that detection window of waving is determined according to pedestrian detection window.
Specific embodiment
Present invention will be further explained below with reference to the attached drawings and specific embodiments.
The region shot by unmanned plane is larger, and the region of waving with clapping object only accounts for the sub-fraction of entire picture, because This is in order to which whether there is or not wave to act for accurate judgement, it is necessary first to determine a general region.Since hand region is too small, directly into The detection of row hand is infeasible.In view of waving to act the general upper right side for betiding target, the invention proposes one kind based on fortune The detection method of waving of dynamic history image.
A pedestrian detector is trained first, and different pedestrians are shot by the camera on unmanned plane, are adopted altogether Collection is no less than the pedestrian image of 500 width as positive template, and the pedestrian's number for participating in acquisition is no less than 100 people.Then from network or Other various databases of person collect it is various do not include pedestrian 200 width more than image as negative sample.Not by collected collection Pedestrian image less than 500 width normalizes to the image that size is 108*36, selects classical histograms of oriented gradients (HOG) special It levies extracting method and feature extraction is carried out to positive negative sample, be trained using svm, obtain a pedestrian detector.
When unmanned plane carries out target with clapping, pedestrian detection, pedestrian detection process are carried out to the video image of shooting first It uses the mode of sliding window (window size 108*36) to carry out, extracts the HOG feature of image in frame to be checked, examined by pedestrian Survey device classify, obtain one whether be pedestrian score, if the score be greater than 0.7, using this frame to be detected as candidate frame. When there are multiple candidate frames, pedestrian candidate frame is ranked up according to classifier score height.Choose wherein that score is most first Object of the high pedestrian candidate frame as detection of waving for the first time, region of waving to it carries out waving to examine by motion history image It surveys, if detection has behavior of waving, testing process of entirely waving terminates, and starts subsequent trace routine；If highest scoring Pedestrian candidate frame do not detect the behavior of waving, then the pedestrian candidate frame of score second is carried out waving to detect, waves to detect Method is identical with the detection method of waving for test object of waving for the first time；It is regular according to this, until detecting some pedestrian candidate frame It has carried out waving to detect but until behavior of not waving in the presence of the behavior of waving or all pedestrian candidate frames.
Next the specific steps of detection proposed by the present invention of waving are introduced:
The window for detection of waving is determined according to the window of pedestrian detection first, as shown in figure 3, the window for detection of waving is big Small is 36*36, positioned at the upper left side of pedestrian detection window.Two windows are the left top of detection window and pedestrian detection window of waving Point differs 12 pixels in x, y-axis respectively.
The video image for remembering unmanned plane shooting is { Pn(x, y) | n=1,2 ... N }.Unmanned plane shooting video image be by Time alignment, what each width image in video image here indicated is first frame, the second frame ... nth frame image.
To each width image in video image, all only retain the pixel value waved in detection window region, even waving to examine The all pixels value for surveying window exterior domain is set to 0.
Since n-th (n >=2) frame image, motion detection is carried out using 3 frame differences, calculation formula is as follows:
Dn(x, y)=Pn-1(x,y)-2Pn(x,y)+Pn+1(x,y) (1)
Then use classical big law (otsu) by Dn(x, y) binaryzation obtains two-value motion information An(x,y)。
The following present invention proposes a kind of calculation method of motion history image, is allowed to preferably characterize the direction letter of movement Breath.Motion history image Hn(x, y) is a kind of global description's method about movement, wherein the value of each pixel is the picture The move time information of vegetarian refreshments, is defined as follows:
As can be seen from the above equation, the pixel grey scale change information of motion history image embodies the direction of behavior campaign, most The pixel brightness value closely moved is maximum, and moves more long pixel and then removed.So motion history image embodies well The space characteristics and temporal information of movement.
The gradient direction namely angle matrix θ of each pixel of motion history image are calculated using sobel operatorn(x, y):
Wherein It is volume Product operation.
It is likely to be obtained the direction vector of mistake on boundary pixel, therefore, before calculating using formula 3, first determines whether to move Whether each pixel on history image is boundary pixel, specific practice are as follows: on motion history image, with any pixel (x0,y0) centered on, one shares 9 pixels in the field that size is 3 × 3, only when 9 pixels are not zero, just thinks picture Element (x0,y0) it is not boundary pixel；Otherwise it is assumed that the point is boundary pixel.When calculating gradient direction, if pixel is boundary picture Element does not have to then calculate that its gradient direction is directly set to 0；If pixel is not boundary pixel, calculated using formula 3. It in this way can be to avoid boundary pixel bring erroneous effects.
The range for the angle matrix being calculated is 0 to 359, in order to simplify calculate, use size for 36 histogram.According to It is secondary to angle matrix θn(x, y) each pixel is traversed, statistic histogram.Then angle corresponding to the maximum value by histogram Spend the movement principal direction ω as the imagen。
Image { the P come is transmitted through from video endn(x, y) | n=1,2 ... N in the second frame image start, according to above-mentioned side Method successively calculates the movement principal direction { ω of each imagen| n=2 ..., N }.N generally takes 30.Then a direction index is calculated Value:
Wherein movement principal direction is judged as that the right side is waved between 46 degree and 134 degree；Principal direction is moved in 226 degree and 314 degree Between, it is judged as that a left side is waved.
When non-zero value reaches 15 in current 30 width image, present invention determine that overall flow terminates there are waving motion. If current pedestrian is determined there is no waving motion, next pedestrian is carried out waving to examine according to the score of pedestrian detection It surveys, until detecting waving motion or whole pedestrians do not wave.

Claims (4)

1. a kind of detection method of waving based on motion history image, which comprises the following steps:
S1 trains a pedestrian detector；
Positive template is acquired, and is normalized；Negative sample is collected later；Then choice direction histogram of gradients (HOG) is special It levies extracting method and feature extraction is carried out to positive negative sample, be trained using svm；
S2 carries out pedestrian detection when unmanned plane carries out target with clapping, to the video image taken, high according to classifier score It is low that pedestrian candidate frame is ranked up, choose object of the pedestrian candidate frame of wherein highest scoring as detection of waving for the first time；
Pedestrian detection is carried out by the way of sliding window；
S3 carries out detection zone determination of waving to the highest pedestrian candidate frame of present score, then passes through fortune in detection zone of waving Dynamic history image carries out waving to detect, which is characterized in that
A detection window of waving is determined as the window of pedestrian detection according to the highest pedestrian candidate frame of present score first, Detection window size of waving is 36*36, positioned at the upper left side of pedestrian detection window；It waves detection window and pedestrian detection window Left vertex differs 12 pixels in x, y-axis respectively；
The video image for remembering unmanned plane shooting is { Pn(x, y) | n=1,2 ... N }, to each width image in video image, all only protect The pixel value waved in detection window region is stayed, even all pixels value for detection window exterior domain of waving is set to 0；
Since n-th (n >=2) frame image, the motion detection of n-th (n >=2) frame image is carried out using 3 frame differences, calculation formula is such as Under:
Dn(x, y)=Pn-1(x,y)-2Pn(x,y)+Pn+1(x,y) (1)
Wherein Pn(x,y)、Pn-1(x,y)、Pn+1(x, y) respectively indicates n-th frame image, the (n-1)th frame image, the (n+1)th frame image；
Then use classical big law by Dn(x, y) binaryzation obtains two-value motion information An(x,y)；
Next it carries out waving to detect by motion history image waving detection zone, the method is as follows:
A. motion history image H is definedn(x, y), wherein the value of each pixel is the move time information of the pixel, it is fixed Justice is as follows:
B. on motion history image, with any pixel point (x0,y0) centered on, one shares 9 pictures in the field that size is 3 × 3 Vegetarian refreshments just thinks pixel (x when 9 pixels are not zero0,y0) it is not boundary pixel；Otherwise it is assumed that the pixel (x0,y0) it is boundary pixel；
When calculating gradient direction, if pixel is boundary pixel, its gradient direction is directly set to 0；If pixel is not Boundary pixel then calculates its gradient direction namely angle matrix θ using sobel operatorn(x, y), as follows:
Wherein It is convolution fortune It calculates；
C. the range for the angle matrix being calculated is 0 to 359, use size for 36 histogram, successively to angle matrix θn (x, y) each pixel is traversed, statistic histogram, then using angle corresponding to the maximum value of histogram as the image Movement principal direction ωn,
D. image { the P come is transmitted through from video endn(x, y) | n=1,2 ... N in the second frame image start, according to the method described above Successively calculate the movement principal direction { ω of each imagen| n=2 ..., N }, then calculate a direction index value:
Wherein movement principal direction is judged as that the right side is waved between 46 degree and 134 degree；Move principal direction 226 degree and 314 degree it Between, it is judged as that a left side is waved；
When non-zero value reaches N/2 in current N width image, then determine that there are waving motions, otherwise, it is determined that not wave to transport It is dynamic；
When detecting waving motion, overall flow terminates S4；It waves if the pedestrian candidate frame of highest scoring does not detect Behavior then carries out waving to examine to the pedestrian candidate frame of score second in the classifier score height ranking results of pedestrian detection in S2 It surveys, detection method of waving is identical with the detection method of waving for test object of waving for the first time；It is regular according to this, until detecting some There is behavior of waving in pedestrian candidate frame or all pedestrian candidate frames have carried out waving to detect behavior of not waving still Until.
2. the detection method of waving according to claim 1 based on motion history image, which is characterized in that in S1, pedestrian The training method of detector is: being shot by the camera on unmanned plane to different pedestrians, acquisition is no less than 500 width altogether Pedestrian image as positive template, the pedestrian's number for participating in acquisition is no less than 100 people；
Then from network or other various databases collect it is various do not include pedestrian 200 width more than image as negative sample This；
Collected pedestrian image is normalized into the image that size is 108*36, choice direction histogram of gradients feature extraction side Method carries out feature extraction to positive negative sample, is trained using svm, obtains a pedestrian detector.
3. the detection method of waving according to claim 1 or 2 based on motion history image, which is characterized in that in S2, When unmanned plane carries out target with clapping, pedestrian detection is carried out to the video image of shooting first, pedestrian detection process uses sliding window The size of mouth is 108*36, extracts the HOG feature of image in frame to be checked, is classified by pedestrian detector, obtaining one is The no score for pedestrian, if the score is greater than 0.7, using this frame to be detected as candidate frame；When there are multiple candidate frames, according to Classifier score height is ranked up pedestrian candidate frame, and the pedestrian candidate frame for choosing wherein highest scoring is used as waves to examine for the first time The object of survey.
4. the detection method of waving according to claim 1 based on motion history image, which is characterized in that in step S3, N Value is more than or equal to 30.
CN201610859376.7A 2016-09-28 2016-09-28 A kind of detection method of waving based on motion history image Active CN106446837B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610859376.7A CN106446837B (en) 2016-09-28 2016-09-28 A kind of detection method of waving based on motion history image

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610859376.7A CN106446837B (en) 2016-09-28 2016-09-28 A kind of detection method of waving based on motion history image

Publications (2)

Publication Number Publication Date
CN106446837A CN106446837A (en) 2017-02-22
CN106446837B true CN106446837B (en) 2019-05-07

Family

ID=58170765

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610859376.7A Active CN106446837B (en) 2016-09-28 2016-09-28 A kind of detection method of waving based on motion history image

Country Status (1)

Country Link
CN (1) CN106446837B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2018176426A1 (en) * 2017-03-31 2018-10-04 深圳市大疆创新科技有限公司 Flight control method for unmanned aerial vehicle, and unmanned aerial vehicle
CN109344899B (en) * 2018-09-30 2022-05-17 百度在线网络技术（北京）有限公司 Multi-target detection method and device and electronic equipment

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102609720A (en) * 2012-01-31 2012-07-25 中国科学院自动化研究所 Pedestrian detection method based on position correction model
CN102789568A (en) * 2012-07-13 2012-11-21 浙江捷尚视觉科技有限公司 Gesture identification method based on depth information
CN102799855A (en) * 2012-06-14 2012-11-28 华南理工大学 Video-streaming-based hand positioning method

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10168785B2 (en) * 2015-03-03 2019-01-01 Nvidia Corporation Multi-sensor based user interface

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102609720A (en) * 2012-01-31 2012-07-25 中国科学院自动化研究所 Pedestrian detection method based on position correction model
CN102799855A (en) * 2012-06-14 2012-11-28 华南理工大学 Video-streaming-based hand positioning method
CN102789568A (en) * 2012-07-13 2012-11-21 浙江捷尚视觉科技有限公司 Gesture identification method based on depth information

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
A Novel Palm Line Extraction and Matching For Personal Identification;HosseinSalimi,SaiedFazli;《International Journal of Science, Engineering and Technology Research 》;20160228;第5卷(第2期);第2-3页

Also Published As

Publication number Publication date
CN106446837A (en) 2017-02-22

Similar Documents

Publication Publication Date Title
CN110348319B (en) Face anti-counterfeiting method based on face depth information and edge image fusion
Mathe et al. Dynamic eye movement datasets and learnt saliency models for visual action recognition
CN104123543B (en) A kind of eye movement recognition methods based on recognition of face
Mathe et al. Actions in the eye: Dynamic gaze datasets and learnt saliency models for visual recognition
CN105022982B (en) Hand motion recognition method and apparatus
CN108810620A (en) Identify method, computer equipment and the storage medium of the material time point in video
CN109829467A (en) Image labeling method, electronic device and non-transient computer-readable storage medium
US8498454B2 (en) Optimal subspaces for face recognition
CN107067413B (en) A kind of moving target detecting method of time-space domain statistical match local feature
US20150054824A1 (en) Object detection method, object detection device, and image pickup device
CN107506703A (en) A kind of pedestrian's recognition methods again for learning and reordering based on unsupervised Local Metric
CN104615986B (en) The method that pedestrian detection is carried out to the video image of scene changes using multi-detector
CN106529499A (en) Fourier descriptor and gait energy image fusion feature-based gait identification method
CN103902978B (en) Face datection and recognition methods
CN111027481B (en) Behavior analysis method and device based on human body key point detection
CN110503000B (en) Teaching head-up rate measuring method based on face recognition technology
CN105512618B (en) Video tracing method
CN110991397B (en) Travel direction determining method and related equipment
CN108960142B (en) Pedestrian re-identification method based on global feature loss function
CN103106414A (en) Detecting method of passer-bys in intelligent video surveillance
CN117133057A (en) Physical exercise counting and illegal action distinguishing method based on human body gesture recognition
CN106446837B (en) A kind of detection method of waving based on motion history image
CN108257148B (en) Target suggestion window generation method of specific object and application of target suggestion window generation method in target tracking
CN102148919B (en) Method and system for detecting balls
CN107886060A (en) Pedestrian's automatic detection and tracking based on video

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