Yanzi Jin

Yanzi Jin 金彦孜
I am a Ph.D. candiate of University of Illinois at Chicago(UIC), working with my advisor Professor Jakob Erikkson in the BITS Networked Systems Laboratory. I received my B.S. in Software Engineering from Dalian University of Technology in 2012.

Research Interests

My research interest lies in computer vision and machine learning, especially in video content analysis. I like interdisciplinary topics and digging into different problems step by step.

Please see more details in my resume and LinkedIn. The code of our system and common video processing tools is available at my GitHub.

Publication

Yanzi Jin and Jakob Eriksson. Fully Automatic, Real-Time Vehicle Tracking for Surveillance Video, 14th Conference on Computer and Robotic Vision, May 2017 for oral presentation. [.pdf]

Dataset

Our dataset containing 13 5-mins long videos is now publicly available with ground truth. The format of the ground truth is "object_id x y width height frame_id if_lost if_occluded if_interpolated label", where the if_interpolated comes from the annotation tool and may not be useful.

Please cite our paper if the dataset is used in your experiment.

@article{yanziVehicleTracker,
 title={Fully Automatic, Real-Time Vehicle Tracking for Surveillance Video},
 author={Jin, Yanzi and Jakob, Eriksson},
 journal={Computer and Robotic Vision},
 year={2017},
}

Current Project

Project Overview

The project is in collaboration with the Illinois Department of Transportation (IDOT), aiming to generate traffic counting from low-resolution surveillance videos. The goal is to realize 24/7 counting despite camera view, weather and other environment factors. The project covers a wide range of problems in computer vision, from high-level object detection/tracking to low-level image processing. Challenges come from the poor quality of the given videos and various camera angles, lighting conditions.

Workflow

The figure below gives the workflow of the system framework. We first apply a background model called ViBe to extract moving objects. After proper noise elimination, we initialize the objects and track them until leaving. This step will generate a trajectory file, which will be passed to counting step. Among those steps, the most important part is the decision of object entering and leaving.

Yanzi Jin

The following is the summary of my current work so far.

  • Designed and implemented a framework for vehicle counting in transportation videos, in collaboration with Illinois Department of Transportation (IDOT).
  • Realized fully automatic vehicle initialization and tracking in surveillance videos, despite camera view, vehicle size and speed, and noise.
  • Came up with an efficient reverse tracking solution to tracking inaccuracy in low-resolution videos and simple occlusion scenarios, extended from ViBe and LK tracker.
  • Implemented an integrated GUI application of the vehicle counting pipeline, for practical use in IDOT.

Future Work

With our framework working well on videos at during day time, we are currently working on the improvement of more challenging cases, which includes:
  • Extreme weather and night videos.
  • Sudden background change due to auto-exposure and video compression.
  • Inaccurate optical flow estimation near image boundary.

Demos

The videos below give a brief idea of my work so far. The first video is the visualization of tracking and counting step, while the last two are the demo of our GUI application.

Publicity

Teaching

  • 2012 Fall: CS111 Program Design
  • 2013 Spring: CS450 Intro to Networking

Contact Information

Address PhoneSorted ascending Internet
University of Illinois at Chicago
Department of Computer Science
Room 1120 SEO (M/C 152)
Chicago, IL 60607
Lab: (312) 413-2103 Email: yjin25@uic.edu

-- Main.yjin25 - 2013-04-16

Topic attachments
I Attachment Action Size Date Who Comment
PDFpdf CV_Yanzi_UIC_CS.pdf manage 148.3 K 2018-02-09 - 23:02 UnknownUser  
PNGpng jenny_profile.png manage 7752.8 K 2017-04-07 - 20:30 UnknownUser  
PDFpdf tracker_CRV17.pdf manage 2713.0 K 2017-04-07 - 20:36 UnknownUser  
PNGpng vehicle_counts.png manage 160.0 K 2015-10-13 - 20:17 UnknownUser  
 
Copyright 2016 The Board of Trustees
of the University of Illinois.webmaster@cs.uic.edu
WISEST
Helping Women Faculty Advance
Funded by NSF