A l e x a n d e r P a t t e r s o n I V
Ph.D. Candidate
UPenn Vision Group
GRASP Laboratory
Department of Computer and Information Science
University of Pennsylvania
Advisor: Kostas Daniilidis
xandey (at) gmail.com
Current Interests:
My main research topic is laser scan data. LiDAR contains a very different sort of data from traditional images, and presents specific challenges whch are not always present in other imaging modalities. I have interest in the fields of object recognition, learning, registration between image and LiDAR data, and data fusion between LiDAR and other modalities. I have been contributing to this field since 2006.
Automatic Alignment of a Camera with a Line Scan LIDAR System
Oleg Naroditsky, Alexander Patterson IV, Kostas Daniilidis
IEEE International Conference on Robotics and Automation, May 2011.
Download: [PDF] [CODE V1.0]Object Detection from Large-Scale 3D Datasets using Bottom-up and Top-down Descriptors
Alexander Patterson IV, Philippos Mordohai, Kostas Daniilidis
European Converence on Computer Vision, October 2008.
Download: [PDF]Fully Automatic Registration of 3D Point Clouds
Ameesh Makadia, Alexander Patterson IV, Kostas Daniilidis
IEEE Conference on Computer Vision and Pattern Recognition, New York, June 2006.
Oral Presentation (CVPR06 accepted 6% of submitted papers for oral presentations)
Download: [PDF]Digitizing Archaeological Excavations from Multiple Views
Xenophon Zabulis, Alexander Patterson IV, Kostas Daniilidis
International Conference in 3D Digital Imaging and Modeling, June 2005.
Download: [PDF]
ARM-S: A realtime Robotic manipulation project which leverages a Barett and accompanying sensor package. We replaced government calibration software and built custom object recognition routines leveraging stereo images, flash lidar, and audio sensors.
Complex Mapping: A realtime Robotic mapping application which maps and fuses laserscan/color image data into 3D map for the robot's controller. Problems include range scanner calibration, image to pointcloud registration/calibration, fast stereo computation and fusion, pose filtering, and super resolution.
Urban Reasoning and Geospatial Exploitation Technology (URGENT): Laser scanner mapping project. Main topic was object recognition, but the goal was to build a system capable of city scale processing with semantic object classification. We implemented segmentation and object recognition components.
Range Image Registration: Use the extended gaussian image to register range images which allows for solving the registration problem using fourier techniques. This is a global approach and serves to initizlize the iterative closest point algorithm.
Archaeoviz: 3D image mapping for achaeology. This is a stereo image capture system which uses surveyed markers and stereo images both to calibrate the cameras and to constrain the stereo search range. We also explored ground penetrating radar processing for archaeology applications.
TeleImmersion: A 3D teleconferencing application originally envisioned as a killer app for Internet2. We produced conceptual demos for various applications of teleconferencing involving stereoscopic display setup, and OpenGL object manipulation and examination. This project lives on at UC Berkeley.
JptView: The Java OpenGL Point Cloud Viewer is intended to visualize pointclouds both inside and outside Matlab, and as an interface to perform manual segmentations. The source is released under the GPL.
I was featured in "Debunked" which aired June 17, 2004. As the video states, we measured how spherical these balls were.
Calib2GL: Calibrated OpenGL camera simulator that I wrote for the Tele-Immersion Project.