IRIS Research Staff

Sreenivas Rangan Sukumar, PhD Student

With the advancement in laser scanning technology, it is now possible to acquire high fidelity 3D data. The 3D spatial information from such scanners serves as a visualization bed for multi-modal surveillance sensors. The idea behind my research is to come up with an automatic data processing framework for the collection of 3D data and integration of multi-sensor information for intelligent visualization. Applications of such a framework targets archival of 3D structures of sensitive environments, virtual models of real world scenes, robotic navigation, inspection and scene analysis.

Surface condition of the runways in an airport is extremely important for proper take off and landing of aircrafts. By surface condition, we are referring to possible cracks and ruts and also debris from previous use and other spurious objects. Thus far, the runway maintenance based on the inspection staff's judgment and experience involves considerable manual intervention and hence the possibility of human error. Our goal for this project is to build a 3D mobile imaging system that can help detect cracks, ruts and objects along the surface accurately and also at a considerable speed compared to the traditional approach. We further propose to integrate video, range and INS information to help archival of the surface condition over time for future wear and tear analysis.

Our approach to shape description burgeons from the seminal work of Shannon. We propose that shapes (of man made objects) can be described as piecewise smooth parts each of which can be uniquely identified using a shape number. We compute the shape number of a surface as the entropy of the curvature density. We segment the parts using a simple region growing procedure and in the future would like to automate this process using the Minima Rule algorithm. Our goal is to describe laser scanned automotive parts and be able to perform similarity search of available CAD models of the automotive components.

Reverse engineering is the process of duplicating an existing component by extracting its physical dimensions. It is almost impossible to get replacements for certain obsolete automobile components. We generate 3D CAD models of such an object of interest by reconstructing the point cloud output of 3D scanners. Our procedure reverse engineers the geometry but not functionality.

The goal of the under vehicle inspection project is to develop mobile robots capable of inspecting a large number of vehicles for concealed nuclear, chemical, and biological weapons. The robot is to be designed with capabilities to capture video, thermal, and range data for threat assessment. Data acquisition is one of the major challenges in this venture and we propose to collect range data from under the vehicle using a sheet of light laser triangulation system. We extend the capability of our sensor system to scan the scene under the vehicle as a color/thermal textured 3D model, suitable for model matching algorithms.