CN108960302A - A kind of head pose estimation method based on random forest - Google Patents

A kind of head pose estimation method based on random forest Download PDF

Info

Publication number
CN108960302A
CN108960302A CN201810638061.9A CN201810638061A CN108960302A CN 108960302 A CN108960302 A CN 108960302A CN 201810638061 A CN201810638061 A CN 201810638061A CN 108960302 A CN108960302 A CN 108960302A
Authority
CN
China
Prior art keywords
head pose
characteristic point
random forest
picture
data
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
CN201810638061.9A
Other languages
Chinese (zh)
Other versions
CN108960302B (en
Inventor
董延超
林敏静
何士波
岳继光
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.)
Tongji University
Original Assignee
Tongji University
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 Tongji University filed Critical Tongji University
Priority to CN201810638061.9A priority Critical patent/CN108960302B/en
Publication of CN108960302A publication Critical patent/CN108960302A/en
Application granted granted Critical
Publication of CN108960302B publication Critical patent/CN108960302B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • 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
    • G06V40/16—Human faces, e.g. facial parts, sketches or expressions
    • G06V40/168—Feature extraction; Face representation
    • G06V40/171—Local features and components; Facial parts ; Occluding parts, e.g. glasses; Geometrical relationships

Landscapes

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

Abstract

The head pose estimation method based on random forest that the present invention relates to a kind of, comprising: step S1: being loaded into plurality of pictures, and imports the characteristic point data and practical head pose data of each picture；Step S2: the picture of all known features point datas He practical head pose data is divided into training set and test set；Step S3: best Random Forest model is obtained using training set picture training, wherein best Random Forest model is used to obtain estimation head pose data according to characteristic point data；Step S4: testing the Random Forest model after training using test set picture, if test error is less than threshold value, S5 is thened follow the steps, if it has not, then return step S1；Step S5: sight estimation is carried out to picture to be estimated using Random Forest model.Compared with prior art, the present invention only needs to utilize 8 characteristic points of face, so that it may carry out the recurrence of each angle of head pose.

Description

A kind of head pose estimation method based on random forest
Technical field
The present invention relates to a kind of head pose estimation methods, estimate more particularly, to a kind of head pose based on random forest Meter method.
Background technique
With the development of science and technology requirement of the people to safety is higher and higher, thus the requirement to face recognition technology also with It is surging.Current many technologies achieve quite satisfactory effect in the environment of the relative ideals such as laboratory, still In practical applications, since the posture of the environment of nature and people is ever-changing, these inevitable factors are all seriously affected The accuracy rate of recognition of face.And for recognition of face and its relevant issues, head pose estimation is an important pre-processing Process, in the case where there is head pose variation, the recognition of face of progress robust is still highly difficult, therefore head pose estimation conduct The prerequisite solved these problems is the important method for improving face recognition technology performance.
In addition, in recent years, it is also higher and higher to the cry of eye tracking research.And in eye tracking, head pose is estimated Meter plays vital role.It is generally believed that usually first head is turned to and is infused when people want to watch some direction attentively Depending on target, then rotate eyes and be placed on target within sweep of the eye, therefore the calculating of sight is using head as reference frame , the estimation of head pose is the important prerequisite of eye tracking.
In conclusion head pose estimation is research important in 21st century computer vision and pattern-recognition neighborhood Content.In recent years, head pose estimation is mainly used in recognition of face, direction of visual lines estimation, and automotive safety auxiliary is driven It sails, because it has various aspects to be widely applied, increasingly causes the attention of researcher.
Summary of the invention
It is an object of the present invention to overcome the above-mentioned drawbacks of the prior art and provide one kind to be based on random forest Head pose estimation method.
The purpose of the present invention can be achieved through the following technical solutions:
A kind of head pose estimation method based on random forest, comprising:
Step S1: being loaded into plurality of pictures, and imports the characteristic point data and practical head pose data of each picture；
Step S2: the picture of all known features point datas He practical head pose data is divided into training set and test Collection；
Step S3: best Random Forest model is obtained using training set picture training, wherein the best random forest mould Type is used to obtain estimation head pose data according to characteristic point data；
Step S4: testing the Random Forest model after training using test set picture, if test error is less than threshold Value, thens follow the steps S5, if it has not, then return step S1；
Step S5: sight estimation is carried out to picture to be estimated using Random Forest model.
The characteristic point data includes:
Two, respectively fisrt feature point and second feature point are arranged in left face characteristic point altogether,
Right face characteristic point is arranged two altogether, symmetrical with left face characteristic point respectively for fourth feature point and third feature,；
Vertical features point is arranged four altogether, is from top to bottom distributed in nose, respectively fifth feature point, the 6th spy vertically Levy point, seventh feature point and eighth feature；
The head pose data include: head pitch angle, azimuth and roll angle.
The step S1 specifically: multiple known features point datas and practical head pose are generated by 3 d modeling software The picture of data.
The step S1 includes:
Step S11: 3 d modeling software generates multiple three-dimensional portraits；
Step S12: three-dimensional portrait characteristic point data and practical head pose data are obtained；
Step S12: plane picture is exported according to three-dimensional portrait.
It is described that best Random Forest model is obtained using the training of training set picture, comprising:
Step S31: the characteristic point data and practical head pose data of training set picture are loaded into；
Step S32: corresponding picture face feature is obtained according to the characteristic point data of each training set picture；
Step S33: face feature and practical head pose data using training set picture are returned using random forest Return analysis, training obtains best Random Forest model, obtains the pass of estimation head pose data according to characteristic point data to find System.
Test process specifically includes in the step S4:
Step S41: the characteristic point data and practical head pose data of test set picture are loaded into；
Step S42: corresponding picture face feature is obtained according to the characteristic point data of each test set picture；
Step S43: using the face feature of test set picture as the input of trained Random Forest model, estimated Head pose data.
The face feature includes:
Lateral length compares feature, comprising:
Wherein: lijFor the distance between i-th characteristic point to jth characteristic point, chalThe average value of preceding 8 length ratio；
Lateral angles compare feature, comprising:
Wherein: ∠ ijk is in the triangle constituted with the i-th characteristic point, jth characteristic point and kth characteristic point, with jth feature Point is the angle value on vertex, cha∠For the average value of preceding 7 angles ratio；
Longitudinal length compares feature, comprising:
Wherein: lijFor the distance between i-th characteristic point to jth characteristic point；
Characteristic point vector and reference axis angle, comprising:
Wherein:For vector lijWith the angle value of the angle of trunnion axis,For vector lijWith the angle of vertical axes Angle value.
Compared with prior art, the invention has the following advantages:
1) it only needs to utilize 8 characteristic points of face, so that it may carry out the recurrence of each angle of head pose.
2) picture that multiple known visual line characteristics amounts and sight legitimate reading are generated by 3 d modeling software, can provide Reliable test set and training set data source.
Detailed description of the invention
Fig. 1 is key step flow diagram of the present invention；
Fig. 2 is characterized selection schematic diagram a little.
Specific embodiment
The present invention is described in detail with specific embodiment below in conjunction with the accompanying drawings.The present embodiment is with technical solution of the present invention Premised on implemented, the detailed implementation method and specific operation process are given, but protection scope of the present invention is not limited to Following embodiments.
A kind of head pose estimation method based on random forest, as shown in Figure 1, comprising:
Step S1: being loaded into plurality of pictures, and imports the characteristic point data and practical head pose data of each picture, wherein such as Shown in Fig. 2, characteristic point data includes:
Left face characteristic point, is arranged two altogether, respectively fisrt feature point and second feature point, and in the present embodiment, left face is special Sign point can choose on the vertex of the left and right of eye socket two,
Right face characteristic point is arranged two altogether, symmetrical with left face characteristic point respectively for fourth feature point and third feature, together It manages, in the present embodiment, right face characteristic point be can choose on right left two vertex of eye socket；
Vertical features point is arranged four altogether, is from top to bottom distributed in nose, respectively fifth feature point, the 6th spy vertically Levy point, seventh feature point and eighth feature；
Head pose data include: head pitch angle, azimuth and roll angle.
Step S1 specifically: multiple known features point datas and practical head pose data are generated by 3 d modeling software Picture, comprising:
Step S11: 3 d modeling software generates multiple three-dimensional portraits；
Step S12: three-dimensional portrait characteristic point data and practical head pose data are obtained；
Step S12: plane picture is exported according to three-dimensional portrait.
Step S2: the picture of all known features point datas He practical head pose data is divided into training set and test Collection；
Step S3: best Random Forest model is obtained using training set picture training, wherein best Random Forest model is used In obtaining estimation head pose data according to characteristic point data, specifically include:
Step S31: the characteristic point data and practical head pose data of training set picture are loaded into；
Step S32: corresponding picture face feature is obtained according to the characteristic point data of each training set picture；
Step S33: face feature and practical head pose data using training set picture are returned using random forest Return analysis, training obtains best Random Forest model, obtains the pass of estimation head pose data according to characteristic point data to find System.
Step S4: testing the Random Forest model after training using test set picture, if test error is less than threshold Value, thens follow the steps S5, if it has not, then return step S1, wherein test process is specifically included:
Step S41: the characteristic point data and practical head pose data of test set picture are loaded into；
Step S42: corresponding picture face feature is obtained according to the characteristic point data of each test set picture；
Step S43: using the face feature of test set picture as the input of trained Random Forest model, estimated Head pose data.
Step S5: sight estimation is carried out to picture to be estimated using Random Forest model.
In the present embodiment, face feature includes:
Lateral length can react the variation in the azimuth direction yaw than feature, this feature, be extracted 10 dimensional features in total, Include:
Wherein: lijFor the distance between i-th characteristic point to jth characteristic point, chalThe average value of preceding 8 length ratio；
Lateral angles can also react the variation in the direction yaw than feature, this feature, be extracted 9 dimension angles in total than feature, Include:
Wherein: ∠ ijk is in the triangle constituted with the i-th characteristic point, jth characteristic point and kth characteristic point, with jth feature Point is the angle value on vertex, cha∠For the average value of preceding 7 angles ratio；
Longitudinal length can react the variation in the direction pitch angle pitch than feature, this feature, comprising:
Wherein: lijFor the distance between i-th characteristic point to jth characteristic point；
Characteristic point vector and reference axis angle, this feature can react the variation in the direction roll angle roll, comprising:
Wherein:For vector lijWith the angle value of the angle of trunnion axis,For vector lijWith the angle of vertical axes Angle value.
, can be while reaching precision by the choosing method of the above face feature, the redundancy for avoiding data from selecting is led The problems such as calculation amount of cause is excessive.

Claims (7)

1. a kind of head pose estimation method based on random forest characterized by comprising
Step S1: being loaded into plurality of pictures, and imports the characteristic point data and practical head pose data of each picture；
Step S2: the picture of all known features point datas He practical head pose data is divided into training set and test set；
Step S3: best Random Forest model is obtained using training set picture training, wherein the best Random Forest model is used According to characteristic point data obtain estimation head pose data；
Step S4: testing the Random Forest model after training using test set picture, if test error is less than threshold value, Step S5 is executed, if it has not, then return step S1；
Step S5: sight estimation is carried out to picture to be estimated using Random Forest model.
2. a kind of head pose estimation method based on random forest according to claim 1, which is characterized in that the spy Levying point data includes:
Two, respectively fisrt feature point and second feature point are arranged in left face characteristic point altogether,
Right face characteristic point is arranged two altogether, symmetrical with left face characteristic point respectively for fourth feature point and third feature,；
Four are arranged altogether, is from top to bottom distributed in nose vertically for vertical features point, respectively fifth feature point, sixth feature point, Seventh feature point and eighth feature；
The head pose data include: head pitch angle, azimuth and roll angle.
3. a kind of head pose estimation method based on random forest according to claim 2, which is characterized in that the step Rapid S1 specifically: the picture of multiple known features point datas He practical head pose data is generated by 3 d modeling software.
4. a kind of head pose estimation method based on random forest according to claim 3, which is characterized in that the step Suddenly S1 includes:
Step S11: 3 d modeling software generates multiple three-dimensional portraits；
Step S12: three-dimensional portrait characteristic point data and practical head pose data are obtained；
Step S12: plane picture is exported according to three-dimensional portrait.
5. a kind of head pose estimation method based on random forest according to claim 2, which is characterized in that described to adopt Best Random Forest model is obtained with training set picture training, comprising:
Step S31: the characteristic point data and practical head pose data of training set picture are loaded into；
Step S32: corresponding picture face feature is obtained according to the characteristic point data of each training set picture；
Step S33: face feature and practical head pose data using training set picture return using random forest and divide Analysis, training obtain best Random Forest model, obtain the relationship of estimation head pose data according to characteristic point data to find.
6. a kind of head pose estimation method based on random forest according to claim 2, which is characterized in that the step Test process specifically includes in rapid S4:
Step S41: the characteristic point data and practical head pose data of test set picture are loaded into；
Step S42: corresponding picture face feature is obtained according to the characteristic point data of each test set picture；
Step S43: using the face feature of test set picture as the input of trained Random Forest model, estimation head is obtained Attitude data.
7. a kind of head pose estimation method based on random forest according to claim 5 or 6, which is characterized in that institute Stating face feature includes:
Lateral length compares feature, comprising:
Wherein: lijFor the distance between i-th characteristic point to jth characteristic point, chalThe average value of preceding 8 length ratio；
Lateral angles compare feature, comprising:
Wherein: ∠ ijk is to be with jth characteristic point in the triangle constituted with the i-th characteristic point, jth characteristic point and kth characteristic point The angle value on vertex, cha∠For the average value of preceding 7 angles ratio；
Longitudinal length compares feature, comprising:
Wherein: lijFor the distance between i-th characteristic point to jth characteristic point；
Characteristic point vector and reference axis angle, comprising:
Wherein:For vector lijWith the angle value of the angle of trunnion axis,For vector lijWith the angle of the angle of vertical axes Value.
CN201810638061.9A 2018-06-20 2018-06-20 Head attitude estimation method based on random forest Active CN108960302B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810638061.9A CN108960302B (en) 2018-06-20 2018-06-20 Head attitude estimation method based on random forest

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810638061.9A CN108960302B (en) 2018-06-20 2018-06-20 Head attitude estimation method based on random forest

Publications (2)

Publication Number Publication Date
CN108960302A true CN108960302A (en) 2018-12-07
CN108960302B CN108960302B (en) 2021-06-04

Family

ID=64490594

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810638061.9A Active CN108960302B (en) 2018-06-20 2018-06-20 Head attitude estimation method based on random forest

Country Status (1)

Country Link
CN (1) CN108960302B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115273180A (en) * 2022-07-01 2022-11-01 南通大学 Online examination invigilating method based on random forest

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102737235A (en) * 2012-06-28 2012-10-17 中国科学院自动化研究所 Head posture estimation method based on depth information and color image
US20150313530A1 (en) * 2013-08-16 2015-11-05 Affectiva, Inc. Mental state event definition generation
CN106529409A (en) * 2016-10-10 2017-03-22 中山大学 Eye ocular fixation visual angle measuring method based on head posture
CN106778579A (en) * 2016-12-07 2017-05-31 电子科技大学 A kind of head pose estimation method based on accumulative attribute
CN108171218A (en) * 2018-01-29 2018-06-15 深圳市唯特视科技有限公司 A kind of gaze estimation method for watching network attentively based on appearance of depth

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102737235A (en) * 2012-06-28 2012-10-17 中国科学院自动化研究所 Head posture estimation method based on depth information and color image
US20150313530A1 (en) * 2013-08-16 2015-11-05 Affectiva, Inc. Mental state event definition generation
CN106529409A (en) * 2016-10-10 2017-03-22 中山大学 Eye ocular fixation visual angle measuring method based on head posture
CN106778579A (en) * 2016-12-07 2017-05-31 电子科技大学 A kind of head pose estimation method based on accumulative attribute
CN108171218A (en) * 2018-01-29 2018-06-15 深圳市唯特视科技有限公司 A kind of gaze estimation method for watching network attentively based on appearance of depth

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115273180A (en) * 2022-07-01 2022-11-01 南通大学 Online examination invigilating method based on random forest
CN115273180B (en) * 2022-07-01 2023-08-15 南通大学 Online examination invigilating method based on random forest

Also Published As

Publication number Publication date
CN108960302B (en) 2021-06-04

Similar Documents

Publication Publication Date Title
CN104268539B (en) A kind of high performance face identification method and system
CN109359526B (en) Human face posture estimation method, device and equipment
CN108171133B (en) Dynamic gesture recognition method based on characteristic covariance matrix
Wang et al. Point cloud and visual feature-based tracking method for an augmented reality-aided mechanical assembly system
CN112084856A (en) Face posture detection method and device, terminal equipment and storage medium
CN113642393B (en) Attention mechanism-based multi-feature fusion sight estimation method
Hernández-Vela et al. BoVDW: Bag-of-Visual-and-Depth-Words for gesture recognition
CN111680678B (en) Target area identification method, device, equipment and readable storage medium
CN105678241B (en) A kind of cascade two dimensional image face pose estimation
CN111209811A (en) Method and system for detecting eyeball attention position in real time
CN113177432A (en) Head pose estimation method, system, device and medium based on multi-scale lightweight network
Ma et al. Correlation filters based on multi-expert and game theory for visual object tracking
CN111681275A (en) Double-feature-fused semi-global stereo matching method
Ao et al. A repeatable and robust local reference frame for 3D surface matching
CN113643329A (en) Twin attention network-based online update target tracking method and system
CN103824063A (en) Dynamic gesture recognition method based on sparse representation
CN109033957B (en) Sight estimation method based on quadratic polynomial
CN113888603A (en) Loop detection and visual SLAM method based on optical flow tracking and feature matching
CN108960302A (en) A kind of head pose estimation method based on random forest
CN113420289A (en) Hidden poisoning attack defense method and device for deep learning model
CN114841965B (en) Steel structure deformation detection method and device, computer equipment and storage medium
CN109658489B (en) Three-dimensional grid data processing method and system based on neural network
CN112990105B (en) Method and device for evaluating user, electronic equipment and storage medium
CN111836072B (en) Video processing method, device, equipment and storage medium
CN116090094A (en) Hull thermal model building method, device and equipment based on infrared thermal imaging

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