Video Object Avoidance Implementation on Embedded Platform

Motion detection is fundamental in various computer vision related applications. In this project, there
are two motion detection techniques being studied, namely optical flow and motion templates. This is
to detect the moving obstacles as well as to classify the direction of the moving obstacles. Optical
flow is the computation to approximate the image motion, while motion templates use the motionhistory-
image (MHI) to keep track of the most recent movement with the timestamp. Besides, this
project also covers the static object detection, where HSV color model classification technique is
used to detect the static obstacles. This technique is based on filtration of color, which depending on
the HSV values of the static objects. Both motion and static detection algorithms will be tested in
Window Visual Studio 2010, before implementing them into the embedded platform, which is
Raspberry Pi. Meanwhile, OpenCV is used as the computer vision library throughout the project. At
the end of this project object, motion templates is selected as a more suitable motion detection
techniques due to its extra information, which is the angle. The HSV technique can detect the static
objects but limited to the calibrated color only