How to Troubleshoot Apps for the Modern Connected Worker
TRACKING OF PARTIALLY OCCLUDED OBJECTS IN VIDEO SEQUENCES
1. Tracking of Partially Occluded Object in
Video Sequences
PRAVEEN PALLAV (1DS09IS061)
Under the Guidance of
Mr. M.T Gopala Krishna
Associate. Professor, Dept. of ISE
Department of Information Science and Engineering
Dayananda Sagar College of Engineering
2. INTRODUCTION
Occlusion:- Occlusions occur when the view of a moving object is blocked
completely or partially by other objects.
TYPES OF OCCLUSION
Tracking of Partially Occluded Object in
Video Sequences
3. Object tracking in computer vision refers to the task of tracking individual
moving objects accurately from one frame to another in an image sequence.
Problem Statement
The problem faced in tracking can be broadly classified into the following:-
Varying illumination conditions.
Partial occlusion of the objects.
Variation in the shape of target.
Objective
Creating a database for experimental purpose in different environment
indoor and outdoor.
Designing a robust object tracking mechanism in video sequences.
The proposed system is simulated in MATLAB R2013.
This proposed system is experimented on standard database available and
our own database.
Tracking of Partially Occluded Object in
Video Sequences
4. Literature Survey
Patches-based Markov random field model for multiple object tracking
under occlusion by Mingjun Wu , Xianrong Peng and Qiheng Zhang in
2010.
They have proposed a new method to this problem, building
on the patch representation of object appearance. They
formulated multiple objects tracking as classification tasks
which competitively use the appearance models of the
interacting objects.
Tracking of Multiple Objects under Partial Occlusion by Bing Han,
Christopher Paulson, Taoran Lu, Dapeng Wu, Jian Li in 2004.
They proposed an algorithm for multi-object tracking under
occlusion based on a new weighted Kanade-Lucasi-Tomasi
(KLT) tracker.
Tracking of Partially Occluded Object in
Video Sequences
5. PROPOSED ALGORITHM
STEP 1 Read the video sequence from the dataset.
STEP 2 Convert video sequences into a set of frames.
STEP 3 Reserve first twenty frames for background registration
STEP 4 Calculate the difference value using
Df = abs( BG – IM )
Where, BG = Background image
IM = Input frame
And update the difference value using
Df = max(Df ,[ ] ,3)
Tracking of Partially Occluded Object in
Video Sequences
6. STEP 5 Create the Binary Mask and apply morphological operation
using the following function
STEP 6 Based on the binary mask obtained, denoised mask is
calculated and plotted.
STEP 7 Using Lucas Kanade Feature Tracker algorithm, labeling of
region of interest with different color components for different
objects.
STEP 8 If no track point is found in object then create new entry in
database and obtain the coordinates and initialize the dictionary.
STEP 9 Tracking is done based on the entries in the dictionary and
proper output is shown in case of occlusion.
Tracking of Partially Occluded Object in
Video Sequences
BW = bwmorph( BW,'bridge')
8. o The efficiency of the algorithm is verified by considering standard
database and our own database available.
o The technique that are used for detection and tracking is Kanade-Lucas-
Tomasi (KLT) tracker.
o The proposed method is used for multi-object tracking under
occlusion by combining multiple cues(Color, Motion, Features ).
RESULTS AND DISCUSSION
Tracking of Partially Occluded Object in
Video Sequences
9. ORIGINAL IMAGE OCCLUDED IMAGE
DIFFERENCE VALUES WITH RESPECT TO BACKGROUND
Tracking of Partially Occluded Object in
Video Sequences
10. DENOISED IMAGE DENOISED IMAGE
Tracking of Partially Occluded Object in
Video Sequences
TRACKED IMAGE
11. RESULT ANALYSIS
Tracking of Partially Occluded Object in
Video Sequences
TRACKING BASED
ON EXISTING SYSTEM
TRACKING BASED
ON PROPOSED SYSTEM
12. APPLICATIONS
Tracking of Partially Occluded Object in
Video Sequences
Human–computer Interaction (HCI) involves the study, planning, and
design of the interaction between people (users) and computers.
Anomaly detection, also referred to as outlier detection refers to detecting
patterns in a given data set that do not conform to an established normal
behavior. The patterns thus detected are called anomalies.
This project can be used as a real-time traffic surveillance system for the
detection, recognition, and tracking of multiple vehicles in roadway
images.
To count the number of objects in a video, multiple object tracking can be
used.
13. APPLICATIONS (Contd…)
Tracking of Partially Occluded Object in
Video Sequences
In robot navigation, the steering system needs to identify different
obstacles in the path to avoid collision. If the obstacles themselves are
other moving objects then it calls for a real-time object tracking
system.
14. CONCLUSION
The proposed method is used for multi-object tracking under
occlusion by combining multiple cues.
Different color patch for different object.
Takes care of partially occluded images.
Dictionary is initialized when new object is detected.
Automated object detection and tracking.
Tracking of Partially Occluded Object in
Video Sequences