By Topic

Landmark automatic identification from three dimensional (3D) data by using Hidden Markov Model (HMM)

Sign In

Cookies must be enabled to login.After enabling cookies , please use refresh or reload or ctrl+f5 on the browser for the login options.

Formats Non-Member Member
$31 $13
Learn how you can qualify for the best price for this item!
Become an IEEE Member or Subscribe to
IEEE Xplore for exclusive pricing!
close button

puzzle piece

IEEE membership options for an individual and IEEE Xplore subscriptions for an organization offer the most affordable access to essential journal articles, conference papers, standards, eBooks, and eLearning courses.

Learn more about:

IEEE membership

IEEE Xplore subscriptions

5 Author(s)
Niu, J.W. ; Dept. of Logistics Eng., Univ. of Sci. & Technol. Beijing, Beijing, China ; Zheng, X.H. ; Zhao, M. ; Fan, N.
more authors

3D anthropometric data can be obtained more easily than several years ago, but to process 3D data is challenging and landmark identification is one of the most challenges. Traditional methods need to manually paste markers on landmarks of the subjects. It is time-consuming and greatly impacts the efficiency of future data analysis. Numerous methods for landmark identification have been proposed, but none with high efficiency, good robustness and strong adaptability has been found yet. In this paper, an approach combined spin image with Hidden Markov Model (HMM) together to automatically identify and locate facial landmarks from 3D face data was proposed. Two hundred 3D head data of young male Chinese were analyzed. Under Unigraphics software, data noise and redundancy were removed and 3D head models were generated. Spin image was adopted to describe the 3D facial landmark local features. With visual check, we defined our facial landmarks and calculated spin images of the facial landmarks. Subsequently, we created a Hidden Markov Model for each landmark, and then used 90 samples as training data to train the model. At last, 60 heads were used as testing samples. Preliminary results show that spin image is highly efficient and robust to pose and lighting. As for HMM, the maximum landmark recognition rate reached 98.33%. There were two landmark HMM models (right orbitale and left cheilion) successful other two models (sellion and right cheilion) quality in general, and the rest of the models had low recognition rate. Further investigation is necessary to evaluate the feature descriptive ability of spin image in future. There are also several aspects related with HMM need to be investigated, such as the convergence of Baum-Welch algorithm and initialization parameters of B etc. Also enhancing efficiency and robustness and using our method to identify other landmarks of human body segments is worth more investigation.

Published in:

Industrial Engineering and Engineering Management (IE&EM), 2011 IEEE 18Th International Conference on  (Volume:Part 1 )

Date of Conference:

3-5 Sept. 2011