Surface Form Modeling
Detailed and accurate surface mapping is a critical competency to support autonomous construction equipment applications. There are several difficulties in accomplishing this, including: high visual uniformity of surface of interest, low resolution and high cost of illumination-based sensors (e.g. LIDAR).
Shape From Polarization (SFP) techniques utilize the fact that surface orientation influences the polarization of reflected light and that knowledge can be used to evaluate relative orientation of surfaces in an image. SPF can be used in conjunction with stereography to acquire high resolution 3D mapping for a field of view.
Construction equipment is generally moving, so obtaining stereo images at a range of polarizations from a single viewpoint would require 6 cameras which is an impractically high number for various reasons. We were able to reduce the number of cameras to 2 by sequentially rotating the polarization and using Simultaneous Location And Mapping (SLAM) to provide the constraints required to rectify sequential images for SFP processing.
Our team was able to extract high precision depth mapping at high resolution for very difficult surfaces including asphalt and dirt.
​
Co-inventor and R&D team lead, provisional patent application assigned to Topcon Positioning Systems.
© 2019 Donna Kelley