فریبا نصیریان

فریبا نصیریان


تاریخ انتشار : Publish : نسخه قابل چاپ Print

دانشكده فني مهندسي

گروه مهندسي کامپیوتر

 

جلسه دفاعیه پایان نامه جهت دريافت درجه كارشناسي ارشد در رشته مهندسی کامپیوتر

گرايش هوش مصنوعی

 

عنوان:

ارایه روشی هوشمند برای استخراج تعامل انسان-انسان و تشخیص رویداد مبتنی بر داده های لیزر


 

استاد راهنما:

دكتر حسن ختن لو

 

داوران:

دکتر میرحسین دزفولیان

دکتر محرم منصوری زاده

 


نگارش:

فریبا نصیریان

 

زمان: سه شنبه  24 شهریور ماه 94 ، ساعت 11 تا 12:30

مکان: آمفی تئاتر دانشکده مهندسی

چكيده:   

از آنجایی که به مرور ربات ها از محیط آزمایشگاه و کارخانجات خارج و وارد زندگی انسان ها شده اند، تحقیق و پژوهش بر روی ربات هایی که قابلیت تفسیر محیط و ارتباط با انسان ها و کار در محیط های پویا را داشته باشند، از اهمیت برخوردار شده است. روش های متفاوتی برای ارتباط ربات با محیط خارج وجود دارد. در این پژوهش سعی بر آن است که از لیزر دامنه یاب برای دریافت اطلاعات از محیط پویای اطراف استفاده کنیم. لیزرهای دامنه یاب ابزارهایی با سرعت و دقت بالا هستند که از پرتوهای لیزر برای تشخیص فاصله تا یک جسم استفاده می کنند. تشخیص وجود یا عدم وجود تعامل بین انسانها در حوزه های متفاوتی کاربرد دارد، نمونه ای از آن، ربات هایی که هدف از ساخت آنها تعامل با انسانها و حضور در اجتماع و پیمایش مسیر در بین افراد می باشد. در این پژوهش سعی شده تا با استفاده از داده هایی که توسط دو لیزر دامنه یاب بدست آمده اند، تعامل بین انسان ها تشخیص و استخراج شود. برای این منظور ابتدا داده ها، ترازبندی و ادغام شده، سپس با محاسبه مدل پس زمینه و حذف آن و بهره گیری از الگوریتم  خوشه بندی طیفی برای خوشه بندی اطلاعات موجود در دسته های فریم ها و یک شبکه عصبی Stacked Auto Encoder، بعنوان طبقه بند، انسان ها را از اشیاء تشخیص داده و با استفاده از ترکیب فیلتر کالمن و روش های رهگیری مبتنی بر ناحیه، افراد تشخیص داده شده را رهگیری و نهایتا با استفاده از اطلاعات حاصل، به تشخیص و استخراج رویداد و تعامل بین انسان ها پرداخته شد. پس از پیاده سازی روش های ذکر شده در نرم افزار  MATLAB، کارایی روش های ارایه شده بر روی داده های جمع آوری شده توسط دو لیزر دامنه یاب HOKUYO و SICK، آزمایش و  صحت آن ها با داده دهای نشانه گذاری شده توسط خبره ی انسانی راستی آزمایی گردید.

 

واژه­های کلیدی: لیزر دامنه یاب، تشخیص و استخراج تعامل، خوشه بندی طیفی، Stacked Auto Encoder  ، رهگیری مبتنی بر ناحیه


Abstract:

Recently robots have been came out from the laboratories and factories and entered to the human lifes. As the result, performing study and research on robots which can interpret and work in the dunamic environments is important. Different ways are available in the content of robot's communication with its environment. In this research an attempt was done to use Laser Range Finders (LRF) to receive information from the dynamic environment. Laser Range Finders are usefull tools with high accuracy and speed that use laser beams to detect the distance to objects.

Detection of interaction among pedestrians has variety of applications in different fields. An illustration would be the robots which interact with human and their goal is working in society and path planning among human. In this study we tried to use the laser data that measured by two laser range finders to detect the events and interactions among pedestrians. To achieve this goal first, we performed data alignment. After that, omitted the background by calculating the background model and clustered data in the groups of frames by spectural clustering algorithm and classified the clusters using Stacked Auto Encoder neural networks to the classes of human and objects.

To implement human tracking, we used the combination of Kalman filter and region based tracking methods. Finally, we extracted events and interactions using the extracted information in previous stages.

After implementing the explained methods in MATLAB software, the efficiency of the proposed methods was tested by the data wich measured by Hokuyo and SICK laser range finders and the ground truth was annotated by an expert.

 

Key Words: Laser Range Finder, extraction of events and interactions, Spectural clustering, Stacked Auto Encoder, region based tracking.



PERSONAL INFORMATION

Fariba Nasiriyan

 


 house number 14,Ershad street, Shahreza, Isfahan, Iran

 031-53224928     09309777221       

 fariba_nasiriyan@yahoo.com , f.nasiriyan92@basu.ac.ir

Sex Female | Date of birth 15/01/1991 | Nationality Iranian

 

 

 

 

 

 

 

EDUCATION AND TRAINING

 

 

Replace with dates (from - to)

 
 
 

M.Sc.

Artificial intelligence, Computer Engineering Department, Bu-Ali Sina University, Hamadan, Iran

(www.basu.ac.ir)

GPA: 19.19 out of 20 (without the thesis grade)

Thesis Title: An intelligent method for extraction of human - human interaction and event detection based on laser data.

Advisor: Dr. Hassan Khotanlou, Associate Professor

B.Sc.

Hardware engineering, Computer Engineering and IT Department, Hamadan University of Technology, Hamadan, Iran

(www.hut.ac.ir)

GPA: 17.52 out of 20                                          

Thesis Title: Plate detection using the morphological functions in digital image processing.

PERSONAL SKILLS

 

 

Mother tongue(s)

Farsi

 

 

Other language(s)

UNDERSTANDING

SPEAKING

WRITING

Listening

Reading

Spoken interaction

Spoken production

 

English

Intermediate

good

fluent

fluent

fluent

                                                               The language levels are examined with TOEFL iBT exam (may-09-2015)

 

Research interests

 

  Robotics

  Object recognition

  Machine learning

  Deep learning

  Digital image processing

  New aspects related to Artificial intelligence

 

Awards and Achievements

 

  2015     First Place Among M.Sc. Artificial Intelligence Engineering Students at Bu-Ali Sina University.

  2014     Elected Student Choosed by the Office of Brilliant Students of BU-Ali Sina University.

  2013     Entered M.Sc. period of Artificial Intelligence using the regulation of brilliance students     (without Taking the Entrance Exam).

  2013     Honored the Opportunity to Enter the M.Sc. Course in Isfahan University of technology (without Taking the Entrance Exam).

  2013     Honored the Opportunity to Enter the M.Sc. Course in Isfahan University of technology (By Taking the Entrance Exam).

  2013     Honored the Opportunity to Enter the M.Sc. Course in Malek Ashtar University of technology in Isfahan (By Taking the Entrance Exam).

  2013     2nd Place Among B.Sc. Hardware Engineering Students at Hamadan University of technology.

  2008     Top Ranked Student Among all High School Students in Academic Laboratory Competitions in Isfahan State.

 

Computer skills

  Programming Language: C++, MATLAB, VHDL, CLIPS, familiar with JAVA and Mysql.

  Worked on FPGA Spartan 3 at Hamadan University of Technology.

  Software & Tools: MATLAB, ModelSim, CodeVision, Proteus, LogicWork.

  Good knowledge of:  Telecommunications such as EWSD and ZIMENS switches.

                                     The standards of data transmission such as USART.

 

ADDITIONAL INFORMATION

 

 

Publications

Presentations

Projects

Conferences

Seminars

Memberships

Research Experiences

   F. Nasiriyan, H. Khotanlou, “Human Detection in Laser Range Data Using Deep Learning and 3-D Objects”, published at the 7th International Conference on Information Technology and Knowledge, (IKT 2015), IEEE, Iran, 2015.

  F. Nasiriyan, M. Afrasiabi, H. Khotanlou, M. Mansoorizade, “Sparse Connectivity and Sparse Activity Using Sequential Feature Selection in Supervised Learning”, to be submitted as a journal.

  F. Nasiriyan, H. Khotanlou, “Classifying Julia Data Using a Sparse Deep Convolutional Neural Network”, under processing.

  F.Nasiriyan, H. Khotanlou, “Performing Spatial and Temporal Alignment of Multiple Laser Range Finder's Measured Data By Means Of Complex Wavelet Structural Similarity Measure”, under processing.

  Working on the M.Sc. Thesis Titled: “An intelligent method for extraction of human - human interaction and event detection based on laser data” including fusion of data, human detection, human tracking and interaction detection in laser data.

  Worked on Deep Neural Networks including: Self-Taught Learning and Unsupervised Feature Learning, Sparse Coding, Sparse Auto Encoders, Softmax Regression, Feature Extraction Using Convolution and Pooling and so on Using Tutorials of Stanford University Which Are Available On:http://deeplearning.stanford.edu/wiki/index.php/UFLDL_Tutorial.

  Performed the implementation and Simulation of USART Data Transmission Standard Using VHDL Language.

  Worked on Multiple Projects in The Field of AVR Micro controllers using CodeVision Tool.

  Worked on Multiple Projects related to Hardware Simulation in Proteus and Logicwork Environment.

  Performed and did Variety of researches about Human-Robot interaction and Laser Data Applications.

 

References

 

 

 

  Dr. H. Khotanlou, Associate Professor, Department of Computer Engineering, Bu-Ali Sina University, Email: khotanlou@basu.ac.ir

  Dr. M. Mansoorizade, Assistant Professor, Department of Computer Engineering, Bu-Ali Sina University, Email: mansoorm@basu.ac.ir

  Dr. M. Dezfoulian, Assistant Professor, Head of Department of Computer Engineering, Bu-Ali Sina University, Email: dezfoulian@basu.ac.ir