CN107977604A - A kind of hand detection method based on improvement converging channels feature - Google Patents

A kind of hand detection method based on improvement converging channels feature Download PDF

Info

Publication number
CN107977604A
CN107977604A CN201711077703.4A CN201711077703A CN107977604A CN 107977604 A CN107977604 A CN 107977604A CN 201711077703 A CN201711077703 A CN 201711077703A CN 107977604 A CN107977604 A CN 107977604A
Authority
CN
China
Prior art keywords
converging channels
sample
image
detection method
feature
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
CN201711077703.4A
Other languages
Chinese (zh)
Other versions
CN107977604B (en
Inventor
简琤峰
张美玉
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.)
Zhejiang University of Technology ZJUT
Original Assignee
Zhejiang University of Technology ZJUT
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 Zhejiang University of Technology ZJUT filed Critical Zhejiang University of Technology ZJUT
Priority to CN201711077703.4A priority Critical patent/CN107977604B/en
Publication of CN107977604A publication Critical patent/CN107977604A/en
Application granted granted Critical
Publication of CN107977604B publication Critical patent/CN107977604B/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
    • G06V40/10—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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
    • G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
    • G06F3/017—Gesture based interaction, e.g. based on a set of recognized hand gestures
    • 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
    • G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Multimedia (AREA)
  • Human Computer Interaction (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Computation (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Image Analysis (AREA)

Abstract

The present invention relates to a kind of based on the hand detection method for improving converging channels feature, by some picture construction data sets collected, converging channels feature is improved, training simultaneously exports Xgboost models, as detection module；After gathering image to be detected, based on improved converging channels feature and obtained Xgboost models are trained, image is detected, obtains hand images therein.The present invention is as the hand detection based on computer vision, with more preferable promotion potential and value, cost is low, more convenient, in combination with improvement ACF, color space complexion model, Edge Boxes and Xgboost, so as to improve the accuracy rate of detecting system, precision is high, detection performance is good.

Description

A kind of hand detection method based on improvement converging channels feature
Technical field
The present invention relates to for reading or identifying printing or written character or for identifying figure, for example, the side of fingerprint The technical field of method or device, it is more particularly to a kind of to contribute to human-computer interaction to gather with what computer vision field was developed based on improvement Close the hand detection method of channel characteristics.
Background technology
In field of human-computer interaction, human-computer interaction commonly relies on the progress of the equipment such as keyboard, mouse, touch-screen, is set with foregoing It is standby in contrast, detected using hand, can substantially reduce the intelligence such as computer by the action of hand to carry out human-computer interaction Equipment uses threshold, and possesses the flexibility of higher.
Hand detection can be divided into two kinds according to data acquisition modes, be to be based on data glove and regarded based on computer respectively Feel.The particular device gathered data of user's wearing is utilized based on data glove, higher precision can be reached；And based on calculating Machine vision is then to utilize camera collection image data, then passes through the analysis to view data, processing detection task.From facility Property and cost angle consider, since the particular devices such as data glove generally require the cost of higher, and the wearing of particular device can Uncomfortable experience can be caused to the process of human-computer interaction, so in general, the hand detection based on computer vision has More preferable promotion potential and value.
In addition, the hand detection method based on computer vision, it can be divided into again according to the different of processing dimension Two kinds of three peacekeepings two dimension.Three-dimensional hand detection method generally uses more mesh cameras, obtains the depth information of image, is obtained with this The characteristic information of more horn of plenty is obtained, however, such method has the problem of modeling difficulty is high, computationally intensive, hardware cost is high；And The hand detection method of two dimension obtains data only with monocular cam, using with high accuracy, the data processing method of performance, Realize higher detection result, however, it is existing based on the hand detection recognition method of two-dimension computer vision there are precision it is low, The problem of poor performance, especially in the picture there are in noise and background there are class colour of skin object in the case of.
The content of the invention
It is a kind of based on improving converging channels features it is a primary object of the present invention to overcome the deficiencies of the prior art and provide Hand detection method.
It is the technical scheme is that a kind of based on the hand detection method for improving converging channels feature, the hand inspection Survey method comprises the following steps：
Step 1.1：Some images are gathered using camera, build data set；
Step 1.2：Data set based on step 1.1, improves converging channels feature；Train and export Xgboost models；
Step 1.3：Image to be detected is gathered using camera；
Step 1.4：The Xgboost models obtained based on the improved converging channels feature of step 1.2 and training, detecting step 1.3 Image to be detected, obtain hand images therein, repeat step 1.3.
Preferably, in the step 1.1, some images are the image for including hand.
Preferably, in the step 1.1, some images at least possess the following conditions：Comprising in picture noise, image background It is that can cover any pixel of all images more than 300LX, hand position there are class skin pixel, illumination condition covering light intensity.
Preferably, in the step 1.1, the hand of some images has can outlining.
Preferably, in the step 1.2, following steps are specifically included：
Step 1.2.1：Data set based on step 1.1, obtains the sample of image, hands over simultaneously ratio by calculating, judges to define sample This is positive sample or negative sample；
Step 1.2.2：The image of data set is converted into YCbCr patterns and HSV patterns, obtains Cb-Cr components plane and H respectively Component；
Step 1.2.3：Using structuring edge detection algorithm, the marginal probability figure of the image of data set is obtained, and utilizes edge Probability graph obtains the marginal probability histogram in 6 directions；
Step 1.2.4：The edge that the Cb-Cr components plane and H components and step 1.2.3 obtained using step 1.2.2 is obtained is general Rate figure and marginal probability histogram improve converging channels feature；
Step 1.2.5：Utilize improved converging channels feature calculation positive sample and the feature vector of negative sample, training Xgboost graders；
Step 1.2.6：Xgboost graders are trained several times, export Xgboost models.
Preferably, in the step 1.2.1, when handing over and ratio is more than 0.5, judge to define sample for positive sample, work as friendship And ratio be less than 0.3 when, judge define sample for negative sample.
Preferably, it is further comprising the steps of in the step 1.2.6：
Step 1.2.6.1：Candidate window is extracted from the image of data set using Edge Boxes algorithms, according to handing over and compare standard Candidate window is divided into positive sample and negative sample；
Step 1.2.6.2：Use the positive sample of the improved converging channels feature calculation step 1.2.6.1 obtained in step 1.2.4 The feature vector of sheet and negative sample；
Step 1.2.6.3：Further with the feature vector training step 1.2.5's of the positive sample of step 1.2.6.2 and negative sample Xgboost graders, export Xgboost models.
Preferably, in the step 1.2.6.1, when handing over and ratio is more than 0.7, judge to define sample for positive sample, when Hand over and ratio be less than 0.5 when, judge define sample for negative sample.
Preferably, in the step 1.4, detection comprises the following steps：
Step 1.4.1：The Xgboost models obtained using the improved converging channels feature of step 1.2 and training, obtain step The Cb-Cr components plane, H components, marginal probability figure and marginal probability histogram of 1.3 image to be detected；
Step 1.4.2：Candidate window is obtained from marginal probability figure using Edge Boxes algorithms, utilizes Cb-Cr component planes The polychrome color space complexion model formed with H components, removes the candidate window not comprising Skin Color Information；
Step 1.4.3：The corresponding feature vector of remaining candidate window is calculated, using the Xgboost models trained and exported, is commented Estimate each feature vector, obtain classification results；
Step 1.4.4：According to classification results, result is demarcated for positive candidate window with rectangle frame in the input image；
Step 1.4.5：Using non-maxima suppression algorithm, merge positive candidate window, obtain detection window to the end, repeat to walk Rapid 1.3.
The present invention provide a kind of optimization based on the hand detection method for improving converging channels feature, if by collecting Dry picture construction data set, improves converging channels feature, training simultaneously exports Xgboost models, as detection module；Gather to be checked After altimetric image, based on improved converging channels feature and obtained Xgboost models are trained, image is detected, obtains it In hand images.The present invention has more preferable promotion potential and value, cost as the hand detection based on computer vision It is low, more convenient, in combination with improvement ACF, color space complexion model, Edge Boxes and Xgboost, so as to improve The accuracy rate of detecting system, precision is high, detection performance is good.
Brief description of the drawings
Fig. 1 is improvement ACF feature schematic diagrames in the present invention；
Fig. 2 is the flow chart of the training of step 1.2 in the present invention；
Fig. 3 is the flow chart of the detection of step 1.4 in the present invention.
Embodiment
To make the purpose, technical scheme and advantage of the embodiment of the present invention clearer, to the technology in the embodiment of the present invention Scheme is clearly and completely described, it is clear that and described embodiment is part of the embodiment of the present invention, rather than whole Embodiment, based on the embodiments of the present invention, those of ordinary skill in the art institute on the premise of creative work is not made The every other embodiment obtained, belongs to the scope of protection of the invention.
The present invention relates to a kind of corresponding more based on the hand detection method for improving converging channels feature, including generation image Layer channel image, is related to ACF（Aggregate Channel Feature）, polychrome color space complexion model, marginal probability figure Deng then using Edge Boxes algorithms extraction candidate window, finally by Xgboost（eXtreme Gradient Boosting）Candidate window is differentiated.
The hand detection method comprises the following steps.
Step 1.1：Some images are gathered using camera, build data set.
In the step 1.1, some images are the image for including hand.
In the step 1.1, some images at least possess the following conditions：Comprising there are class in picture noise, image background Skin pixel, illumination condition covering light intensity are that can cover any pixel of all images more than 300LX, hand position.
In the step 1.1, the hand of some images has can outlining.
In the present invention, image includes picture noise, there are class colour of skin object, such as face, and illumination condition and hand institute Different in position, this is to make trained data set, and the field faced is needed when being applied to detection comprising method as much as possible Scape.
Step 1.2：Data set based on step 1.1, improves converging channels feature；Train and export Xgboost models.
In the step 1.2, following steps are specifically included.
Step 1.2.1：Data set based on step 1.1, obtains the sample of image, hands over simultaneously ratio by calculating, it is fixed to judge Adopted sample is positive sample or negative sample.
In the step 1.2.1, when handing over and ratio is more than 0.5, judge to define sample for positive sample, when friendship and ratio is small When 0.3, judge to define sample for negative sample.
Step 1.2.2：The image of data set is converted into YCbCr patterns and HSV patterns, Cb-Cr components is obtained respectively and puts down Face and H components.
Step 1.2.3：Using structuring edge detection algorithm, the marginal probability figure of the image of data set is obtained, and is utilized Marginal probability figure obtains the marginal probability histogram in 6 directions.
Step 1.2.4：The side that the Cb-Cr components plane and H components and step 1.2.3 obtained using step 1.2.2 is obtained Edge probability graph and marginal probability histogram improve converging channels feature.
Step 1.2.5：Utilize improved converging channels feature calculation positive sample and the feature vector of negative sample, training Xgboost graders.
Step 1.2.6：Xgboost graders are trained several times, export Xgboost models.
It is further comprising the steps of in the step 1.2.6.
Step 1.2.6.1：Candidate window is extracted from the image of data set using Edge Boxes algorithms, according to handing over and compare Candidate window is divided into positive sample and negative sample by standard.
In the step 1.2.6.1, when handing over and ratio is more than 0.7, judge to define sample for positive sample, when friendship and ratio During less than 0.5, judge to define sample for negative sample.
Step 1.2.6.2：Use the improved converging channels feature calculation step 1.2.6.1's obtained in step 1.2.4 The feature vector of positive sample and negative sample.
Step 1.2.6.3：Further with the positive sample of step 1.2.6.2 and the feature vector training step of negative sample 1.2.5 Xgboost graders, export Xgboost models.
In the present invention, structuring Boundary extracting algorithm is similar with structuring random forest, can be defeated using image block as input Go out the probability distribution that all pixels that the neighborhood at center includes belong to edge pixel, this algorithm can use direct Open-Source Tools Case is completed.
In the present invention, improve person before converging channels feature refers to and replace original ACF passages.
In the present invention, hand over and than being the important indicator for weighing relation between testing result and actual result.
In the present invention, testing result is denoted as rectangle T, the actual minimum rectangle comprising hand be rectangle G, then friendship and ratio Calculation formula is, whenIt is bigger, illustrate testing result and the actual minimum rectangle for including hand It is closest.Under normal circumstances, when handing over and compareDuring more than 0.5, it as a result may be considered that and detect successfully.In actual operation When Cheng Zhong, friendship and ratio are less than 0.3, interception image block can be considered as negative sample.
In the present invention, Xgboost graders establish a label from feature vector to result, i.e., whether include hand Mapping.
In the present invention, the candidate window quantity obtained by sliding window technique can be greatly lowered in Edge Boxes, and Significantly improve the performance and precision of hand detection method.But it should be noted that Edge Boxes can only be protected to greatest extent Window of the card comprising hand is appeared in candidate window, it is impossible to ensures that all candidate windows all include hand, so needing root According to the candidate window that it is produced, Xgboost graders are further trained.
In the present invention, the characteristics of due to Edge Boxes algorithms, i.e., have objects of opposite integrity profile can for those With the accuracy of detection with higher, hand included in image all has relatively complete profile.Under normal circumstances, this point It can be realized by articles for daily use such as wrist-watch, bracelet, coat-sleeves.
Step 1.3：Image to be detected is gathered using camera.
Step 1.4：The Xgboost models obtained based on the improved converging channels feature of step 1.2 and training, detecting step 1.3 image to be detected, obtains hand images therein, repeat step 1.3.
In the step 1.4, detection comprises the following steps.
Step 1.4.1：The Xgboost models obtained using the improved converging channels feature of step 1.2 and training, are walked The Cb-Cr components plane, H components, marginal probability figure and marginal probability histogram of rapid 1.3 image to be detected.
Step 1.4.2：Candidate window is obtained from marginal probability figure using Edge Boxes algorithms, utilizes Cb-Cr components The polychrome color space complexion model that plane and H components are formed, removes the candidate window not comprising Skin Color Information.
Step 1.4.3：The corresponding feature vector of remaining candidate window is calculated, uses the Xgboost moulds trained and exported Type, assesses each feature vector, obtains classification results.
Step 1.4.4：According to classification results, result is demarcated for positive candidate window with rectangle frame in the input image.
Step 1.4.5：Using non-maxima suppression algorithm, merge positive candidate window, obtain detection window to the end, weight Multiple step 1.3.
It is almost mutual between the corresponding feature of diverse location in the multi-channel feature image obtained at this time in the present invention Independent.This is the basis that subsequent step is realized.
In the present invention, candidate window is being obtained from characteristic image using Edge Boxes algorithms, and calculate each candidate During the corresponding feature vector of window, since the corresponding feature of diverse location does not interdepend, so candidate window is in spy It is exactly its feature vector to levy corresponding image block on image.
In Cb-Cr components plane of the human body complexion in YCbCr and the H components in HSV, there are very strong aggregation properties, this Mean that Cb-Cr components plane and H components can be used for distinguishing skin pixel and non-skin pixel, improvement ACF of the invention will Include Cb-Cr and H components.
In the present invention, in order to handle the interference produced in background comprising class colour of skin object and noise, this method is drawn to ACF Enter marginal probability histogram, this is because edge graph can retain the side between different objects while noise jamming is omitted Boundary.
In the present invention, marginal probability figure will be obtained by structuring edge detection algorithm.
In the present embodiment, using camera collection image, 10 characteristic images are calculated according to improved ACF algorithms, wherein, Cb-Cr components and H components may be constructed a polychrome color space complexion model；Obtained using Edge Boxes algorithms on edge graph Candidate window is taken, using polychrome color space complexion model, removes those candidate windows for not including Skin Color Information；Calculate each wait Select the corresponding feature vector of window, and use Xgboost graders, assess each feature vector, draw classification results, according to point Class as a result, result is positive candidate window, in the input image by with rectangle frame calibrate come；Using non-maxima suppression, close And these positive candidate windows, obtain detection window to the end.
In the present embodiment, non-maxima suppression, refers to when the friendship between two positive candidate windows and ratio is more than some threshold value When, abandon wicket and retain big window, i.e. non-maximum is filtered out, leave behind most can detection of expression object frame.
The present invention improves converging channels feature, training simultaneously exports by some picture construction data sets collected Xgboost models, as detection module；After gathering image to be detected, obtained based on improved converging channels feature and training Xgboost models, are detected image, obtain hand images therein.The present invention is as the hand based on computer vision Detection, has more preferable promotion potential and value, cost is low, more convenient, in combination with improvement ACF, the color space colour of skin Model, Edge Boxes and Xgboost, so as to improve the accuracy rate of detecting system, precision is high, detection performance is good.

Claims (9)

  1. It is 1. a kind of based on the hand detection method for improving converging channels feature, it is characterised in that：The hand detection method includes Following steps：
    Step 1.1：Some images are gathered using camera, build data set；
    Step 1.2：Data set based on step 1.1, improves converging channels feature；Train and export Xgboost models；
    Step 1.3：Image to be detected is gathered using camera；
    Step 1.4：The Xgboost models obtained based on the improved converging channels feature of step 1.2 and training, detecting step 1.3 Image to be detected, obtain hand images therein, repeat step 1.3.
  2. It is 2. according to claim 1 a kind of based on the hand detection method for improving converging channels feature, it is characterised in that：Institute State in step 1.1, some images are the image for including hand.
  3. It is 3. according to claim 2 a kind of based on the hand detection method for improving converging channels feature, it is characterised in that：Institute State in step 1.1, some images at least possess the following conditions：Comprising there are class skin pixel, light in picture noise, image background It is that can cover any pixel of all images more than 300LX, hand position according to Condition Coverage Testing light intensity.
  4. It is 4. according to claim 3 a kind of based on the hand detection method for improving converging channels feature, it is characterised in that：Institute State in step 1.1, the hand of some images has can outlining.
  5. It is 5. according to claim 1 a kind of based on the hand detection method for improving converging channels feature, it is characterised in that：Institute State in step 1.2, specifically include following steps：
    Step 1.2.1：Data set based on step 1.1, obtains the sample of image, hands over simultaneously ratio by calculating, judges to define sample This is positive sample or negative sample；
    Step 1.2.2：The image of data set is converted into YCbCr patterns and HSV patterns, obtains Cb-Cr components plane and H respectively Component；
    Step 1.2.3：Using structuring edge detection algorithm, the marginal probability figure of the image of data set is obtained, and utilizes edge Probability graph obtains the marginal probability histogram in 6 directions；
    Step 1.2.4：The edge that the Cb-Cr components plane and H components and step 1.2.3 obtained using step 1.2.2 is obtained is general Rate figure and marginal probability histogram improve converging channels feature；
    Step 1.2.5：Utilize improved converging channels feature calculation positive sample and the feature vector of negative sample, training Xgboost graders；
    Step 1.2.6：Xgboost graders are trained several times, export Xgboost models.
  6. It is 6. according to claim 5 a kind of based on the hand detection method for improving converging channels feature, it is characterised in that：Institute State in step 1.2.1, when handing over and ratio is more than 0.5, judge to define sample for positive sample, when handing over and ratio is less than 0.3, sentence It is negative sample to conclude adopted sample.
  7. It is 7. according to claim 5 a kind of based on the hand detection method for improving converging channels feature, it is characterised in that：Institute State in step 1.2.6, it is further comprising the steps of：
    Step 1.2.6.1：Candidate window is extracted from the image of data set using Edge Boxes algorithms, according to handing over and compare standard Candidate window is divided into positive sample and negative sample；
    Step 1.2.6.2：Use the positive sample of the improved converging channels feature calculation step 1.2.6.1 obtained in step 1.2.4 The feature vector of sheet and negative sample；
    Step 1.2.6.3：Further with the feature vector training step 1.2.5's of the positive sample of step 1.2.6.2 and negative sample Xgboost graders, export Xgboost models.
  8. It is 8. according to claim 7 a kind of based on the hand detection method for improving converging channels feature, it is characterised in that：Institute State in step 1.2.6.1, when handing over and ratio is more than 0.7, judge to define sample for positive sample, when handing over and ratio is less than 0.5, Judge to define sample for negative sample.
  9. It is 9. according to claim 1 a kind of based on the hand detection method for improving converging channels feature, it is characterised in that：Institute State in step 1.4, detection comprises the following steps：
    Step 1.4.1：The Xgboost models obtained using the improved converging channels feature of step 1.2 and training, obtain step The Cb-Cr components plane, H components, marginal probability figure and marginal probability histogram of 1.3 image to be detected；
    Step 1.4.2：Candidate window is obtained from marginal probability figure using Edge Boxes algorithms, utilizes Cb-Cr component planes The polychrome color space complexion model formed with H components, removes the candidate window not comprising Skin Color Information；
    Step 1.4.3：The corresponding feature vector of remaining candidate window is calculated, using the Xgboost models trained and exported, is commented Estimate each feature vector, obtain classification results；
    Step 1.4.4：According to classification results, result is demarcated for positive candidate window with rectangle frame in the input image；
    Step 1.4.5：Using non-maxima suppression algorithm, merge positive candidate window, obtain detection window to the end, repeat to walk Rapid 1.3.
CN201711077703.4A 2017-11-06 2017-11-06 Hand detection method based on improved aggregation channel characteristics Active CN107977604B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201711077703.4A CN107977604B (en) 2017-11-06 2017-11-06 Hand detection method based on improved aggregation channel characteristics

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201711077703.4A CN107977604B (en) 2017-11-06 2017-11-06 Hand detection method based on improved aggregation channel characteristics

Publications (2)

Publication Number Publication Date
CN107977604A true CN107977604A (en) 2018-05-01
CN107977604B CN107977604B (en) 2021-01-05

Family

ID=62013079

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201711077703.4A Active CN107977604B (en) 2017-11-06 2017-11-06 Hand detection method based on improved aggregation channel characteristics

Country Status (1)

Country Link
CN (1) CN107977604B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110889387A (en) * 2019-12-02 2020-03-17 浙江工业大学 Real-time dynamic gesture recognition method based on multi-track matching

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103530607A (en) * 2013-09-30 2014-01-22 智慧城市系统服务（中国）有限公司 Method and device for hand detection and hand recognition
CN105930803A (en) * 2016-04-22 2016-09-07 北京智芯原动科技有限公司 Preceding vehicle detection method based on Edge Boxes and preceding vehicle detection device thereof
US20160343146A1 (en) * 2015-05-22 2016-11-24 International Business Machines Corporation Real-time object analysis with occlusion handling
CN106203385A (en) * 2016-07-20 2016-12-07 东方网力科技股份有限公司 A kind of driver's hand-held phone behavioral value method and device
WO2017095543A1 (en) * 2015-12-01 2017-06-08 Intel Corporation Object detection with adaptive channel features
US20170193313A1 (en) * 2016-01-04 2017-07-06 Texas Instruments Incorporated Real time traffic sign recognition
CN107066935A (en) * 2017-01-25 2017-08-18 网易（杭州）网络有限公司 Hand gestures method of estimation and device based on deep learning

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103530607A (en) * 2013-09-30 2014-01-22 智慧城市系统服务（中国）有限公司 Method and device for hand detection and hand recognition
US20160343146A1 (en) * 2015-05-22 2016-11-24 International Business Machines Corporation Real-time object analysis with occlusion handling
WO2017095543A1 (en) * 2015-12-01 2017-06-08 Intel Corporation Object detection with adaptive channel features
US20170193313A1 (en) * 2016-01-04 2017-07-06 Texas Instruments Incorporated Real time traffic sign recognition
CN105930803A (en) * 2016-04-22 2016-09-07 北京智芯原动科技有限公司 Preceding vehicle detection method based on Edge Boxes and preceding vehicle detection device thereof
CN106203385A (en) * 2016-07-20 2016-12-07 东方网力科技股份有限公司 A kind of driver's hand-held phone behavioral value method and device
CN107066935A (en) * 2017-01-25 2017-08-18 网易（杭州）网络有限公司 Hand gestures method of estimation and device based on deep learning

Non-Patent Citations (10)

* Cited by examiner, † Cited by third party
Title
ESHED OHN-BAR ET AL: ""Beyond Just Keeping Hands on the Wheel: Towards Visual Interpretation of Driver Hand Motion Patterns"", 《2014 IEEE 17TH INTERNATIONAL CONFERENCE ON INTELLIGENT TRANSPORTATION SYSTEMS (ITSC)》 *
N. DAS等: ""On Performance Evaluation of Driver Hand Detection Algorithms: Challenges, Dataset, and Metrics"", 《2015 IEEE 18TH INTERNATIONAL CONFERENCE ON INTELLIGENT TRANSPORTATION SYSTEMS》 *
SHIYANG YAN ET AL: ""Multiscale Convolutional Neural Networks for Hand Detection"", 《APPLIED COMPUTATIONAL INTELLIGENCE AND SOFT COMPUTING》 *
Y. XIA等: ""Combination of ACF detector and multi-task CNN for hand detection"", 《2016 IEEE 13TH INTERNATIONAL CONFERENCE ON SIGNAL PROCESSING (ICSP)》 *
李庆武等: ""基于多尺度聚合通道特征的实时行人检测"", 《电子测量与仪器学报》 *
王真等: ""利用多颜色信息融合的自适应肤色建模"", 《计算机工程与应用》 *
维基百科: "《https://en.wikipedia.org/w/index.php?title=Jaccard_index&oldid=788825557》", 3 July 2017 *
维基百科: "《https://en.wikipedia.org/w/index.php?title=XGBoost&oldid=7973352》", 26 August 2017 *
苏松志等编著: "《行人检测：理论与实践》", 31 March 2016, 厦门:厦门大学出版社 *
陈慧岩等编著: "《车辆信息技术》", 31 December 2013, 北京:北京理工大学出版社 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110889387A (en) * 2019-12-02 2020-03-17 浙江工业大学 Real-time dynamic gesture recognition method based on multi-track matching

Also Published As

Publication number Publication date
CN107977604B (en) 2021-01-05

Similar Documents

Publication Publication Date Title
CN103632132B (en) Face detection and recognition method based on skin color segmentation and template matching
CN108319953B (en) Occlusion detection method and device, electronic equipment and the storage medium of target object
Rekha et al. Shape, texture and local movement hand gesture features for indian sign language recognition
CN109101871A (en) A kind of living body detection device based on depth and Near Infrared Information, detection method and its application
CN103034852B (en) The detection method of particular color pedestrian under Still Camera scene
CN103443804B (en) Method of facial landmark detection
CN105989331B (en) Face feature extraction element, facial feature extraction method, image processing equipment and image processing method
CN108256421A (en) A kind of dynamic gesture sequence real-time identification method, system and device
CN102194108B (en) Smile face expression recognition method based on clustering linear discriminant analysis of feature selection
CN111126240B (en) Three-channel feature fusion face recognition method
Yarlagadda et al. A novel method for human age group classification based on Correlation Fractal Dimension of facial edges
CN111582197A (en) Living body based on near infrared and 3D camera shooting technology and face recognition system
CN110728185B (en) Detection method for judging existence of handheld mobile phone conversation behavior of driver
Rao et al. Sign Language Recognition System Simulated for Video Captured with Smart Phone Front Camera.
CN107169479A (en) Intelligent mobile equipment sensitive data means of defence based on fingerprint authentication
CN110956099B (en) Dynamic gesture instruction identification method
CN106200971A (en) Man-machine interactive system device based on gesture identification and operational approach
CN112114675B (en) Gesture control-based non-contact elevator keyboard using method
Vishwakarma et al. Simple and intelligent system to recognize the expression of speech-disabled person
Jambhale et al. Gesture recognition using DTW & piecewise DTW
Cheong et al. A novel face detection algorithm using thermal imaging
CN112101208A (en) Feature series fusion gesture recognition method and device for elderly people
CN106611158A (en) Method and equipment for obtaining human body 3D characteristic information
CN106909883A (en) A kind of modularization hand region detection method and device based on ROS
CN105930793A (en) Human body detection method based on SAE characteristic visual learning

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
GR01 Patent grant
GR01 Patent grant