International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395 -0056
Volume: 04 Issue: 04 | Apr -2017
p-ISSN: 2395-0072
www.irjet.net
Object Tracking in Video Using Flip-Invariant Scale Invariant Feature Transform Sasmit M. Gokhale1, Tanmay K. Deshmukh2, Vinod V. Dalavi3, Prof. Uttara Gogate4 Department of Computer Engineering, Shivajirao S. Jondhale College of Engineering, Dombivli Maharashtra, India. ---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - The Scale Invariant Feature Transform (SIFT) is
Some of the interesting applications of object detection are:-
invariant to rotation, lighting and scale changes in images. Due to this property of SIFT, it is very popular keypoint descriptor. However, it is not invariant to flips, and this is one of the biggest drawbacks of SIFT. This new descriptor, named as flip-invariant SIFT preserves the original properties of SIFT along with being invariant to flips. Flip Invariant SIFT starts by estimating the dominant curl of a local patch and then performs flipping before computation of SIFT. We have implemented F-SIFT for object detection in a video.
1) Face detection- This is one of the most widely used applications of object detection. Its use is not just in mobile phones and other electronic devices but also on the internet, for example, when you upload any image on social networking websites it can also detect faces in the image. 2) Vehicle detection- Object detection with tracking can help us determine the speed with which vehicle is travelling, thus it can be used in departments such as traffic police to catch people who exceed the speed limit.
Key Words: Flip-Invariant Scale Invariant Feature Transform (F-SIFT), Scale Invariant Feature Transform (SIFT), Object Detection, Object Tracking, Frames.
3) People counting- Object detection can be used to count people from a certain image. It is used to count amount of people visiting a store or a certain area.
1. INTRODUCTION
4) Forensics- It can be used to find any objects on the crime scene that could be used as evidence.
There are many applications developed recently that use the Scale Invariant Feature Transform make object classifiers. The main advantages of SIFT [1] it is invariant to rotation, lighting and scaling. It is also invariant to displacement of pixels in a local region. SIFT is computed over a local salient region. The region is rotated to its dominant orientation and is located using multi-scale detection. Thus, the descriptor is invariant to both scaling and rotation. Also SIFT [1] is invariant to color and lighting because of spatial partitioning. SIFT has the above listed merits, but however the main disadvantage is that SIFT is invariant to flip, which means that if the image is flipped, the result of SIFT [1] changes too. Flip is something that can be encountered a lot of times. Flip can occur because of change in point of views. Also, it can be seen of different video platforms like YouTube that people perform horizontal flipping on their videos or images to hide copyright violations. The figure 1 (a) and 1 (b) demonstrate flipping.
Fig -1(a): Original Image
1.1 Object detection and its applications
2. EXISTING SYSTEM
Object detection is a technology that is related to image processing. It is the process of finding objects in any image. Object detection has a lot of scope these days in different fields like forensic to find evidence in photograph of crime scenes and other technologies like fingerprint detection and face detection.
Š 2017, IRJET
|
Impact Factor value: 5.181
Fig -1(b): Flipped Image
The existing detection systems store the video features in the form of codewords. A simple representation of video features makes the system more efficient. Existing system uses Scale Invariant Feature Transform. SIFT [1] is derived from directionally sensitive gradient fields,
|
ISO 9001:2008 Certified Journal
|
Page 1708