Wavy Surface cause problems such as: Blurred image and Reflections of light from the water surface.
The Problem:
Wavy Surface cause problems such as: Blurred image and Reflections of light from the water surface. In addition it is not trivial to detect the moving object- due to the wave movement it seems that the floor moves as well. Also there are random image distortions due to refractions.
Project Goals:
- Automatic Detection of moving Object
- Tracking the Moving Objects
- Triangulate Object’s Position
The Solution
- Polarizers were in use to decrease the reflections from the water surface.
- Automatic Detection of moving Object:
Using Harris Corner Algorithm to get initial points.
Tracking those points using Kernel Tracker (20-30 frames),
Calculating Standard Deviation-STD (in pixels), of the tracked points.
Using an Empiric threshold on the STD Results to get the wanted object.
- Tracking:
Tracking with Locally Orderless Tracker (LOT) using the pixel coordinates which were found at the Automatic Detection stage.
- Triangulate Object’s Position:
Using Caltech’s calibration tool, Snell’s law, trigonometric 2D and 3D calculations we managed to calculate the object’s 3D coordinates. We smoothed the results and corrected errors which occurred due to LOT miss. Furthermore, we account for distortions due to the wave’s movement.
How Kernel Tracker works?
- Mean Shift Problem
- Continue 1 & 2 until convergence
Results
