Multicore DSPs thermal image sequence processing for real time motion detection in thermal surveillance system

Abstract

Motion detection of objects or people is an important operation in visual and also in thermo visual surveillance systems. From speed of execution and accuracy of motion detection operation strongly are depended the main characteristics of a real working surveillance system. Therefore, the goal of this article is to develop of a multicore DSP thermal image sequence processing module suitable for real time objects or people motion detection and tracking and also to present here the experimental results from this development. The development of multicore DSP module is based on Xilinx module Zynq-7000 All Programmable SoC. The proposed algorithm of motion detection in thermal images is first modeled using OpenCV library functions for image processing, and then the algorithm is accelerating using Xilinx Vivado HLS Video Libraries to implement the algorithm of motion detection in thermal images in Xilinx module Zynq-7000. The presented in this article experimental results show the ability to detect in real time motions of objects or people in thermal images and to use the results for next steps of a whole algorithm for thermo visual surveillance - objects and people tracking in thermal images.