Smart Human Object Identification and Tracking on Soc Through Adaptive TRI-Class Thresholding in Real Time Environment

Authors

  • Dr R Dhaya ,Dr R Kanthavel , Dr M Mahalaskhmil

Abstract

— In this paper we intend to incorporate, a method into a real time system that includes embedded
SoC hardware supported with CPU, GPU, and CMOS Image sensor, to acquire real time image. The
ultimate aim of this paper is to achieve enhanced image segmentation by identifying a different object, and
track a particular identified object out of multiple objects from the acquired real time image. In this system,
a combination of Adaptive Tri-class Otsu’s Method threshold selection and Haar classifier methods have
been applied to enhance the real time images with de-noising effect based on a threshold value and
transforming a grayscale image into a different grayscale image. In addition identifying the particular object
out of the multiple objects from the enhanced real time image under various dynamic environments has also
been done. The identification and tracking of the particular object has here been based on a multi-scale
values. The combination of methods implemented on the SoC results an accurate customized identification
and tracking of the particular identified object

Published

2020-04-30

Issue

Section

Articles