Project Goal
The goal of this project is to find and track moving objects on the ground from a camera flown on a low-flying helicopter platform. The following figure displays two images taken with one of Stanford's helicopters
Figure 1: Images of the ground taken with the Stanford Autonomous Helicopter. The left image shows the helicopter's shadow as the sole moving object, whereas the right image depicts a moving person.
For this project, we will attach a downward looking camera on the helicopter to gather a stabilized stream of NTSC video data at frame rate. By tracking moving objects on the ground, we hope to enable the helicopter to increase its awareness of its environment.
Project Scope
This project involves the analysis of images acquired by a moving helicopter platform. Images have to be processed to determine the optical flow. Discontinuities in the optical flow have to be analyzed; if such discontinuities run counter to the main direction of the optical flow, they shall become candidates of moving objects. Such candidates shall be tracked over multiple image frames, to determine reliably whether they correspond to actual moving objects.
Tasks
The project will be accomplished through the following tasks.You will be working closely with the Stanford Autonomous Helicopter Lab, which will provide the necessary infrastructure to collect the data and evaluate your results. The Lab will provide model helicopters flown by one of our experienced pilots.
- Task 1: Instrumentation and calibration: Calibrate existing camera, attach camera on helicopter
- Task 2: Optical flow: Devise a robust algorithm for calculating the optical flow in the image sequence.
- Task 3: Detecting potential moving objects: Within a single optical flow, detect regions whose motion runs counter to the predominant flow direction. Call these candidates of moving objects.
- Task 4: Classification: Track candidates over multiple frames to classify them with high reliability as moving or non-moving..
- Extension: Estimate the relative location of these moving objects.
Project Status
David Lieb, Andrew Lookingbill, David Stavens, and
Point of Contact
Sebastian Thrun
Midterm Report
submitted
Final Report
submitted