I. CLIF
v Columbus Large Image Format
v EO aerial imagery around OSU
campus
v Frame rate: 2 frames per
second
v Typical sensor altitude: 7000
ft (2.1 km)
v Sensor: 2 x 3 array of 6
cameras
II. Goal
v
Track objects such as cars and trucks
III. Alignment
Features: Harris Corner Descriptors: SIFT Descriptor Matching
RANSAC based homography
fitting
IV. Background Subtraction
v
Gradient Suppression
V. Tracking
1. Tracking
Framework: Bipartite Graph Matching
v Nodes at time t and t+1
§ Edges to nodes at t+1
v Occlusion nodes
§ Edges to occlusion nodes
v Dead end tracks from p frames
back
§ Reacquisition edges
v Solve using Hungarian Algorithm
2. Scene
Constraints for Initialization
v Accurate velocity estimate is
needed to obtain proper correspondences which is not available initially
v Use local contextual
information to help with assignment
2a. Global
Velocity
v Compute all possible velocity
orientations between two frames.
v Obtain histogram.
v Select mean of histogram mode
as the global velocity.
v Compute weight as:
2b. Neighbor
Context:
For every object to match:
v Compute vectors from it’s
position to neighboring objects
v Obtain 2D histogram of
orientation and magnitude
v Compute weight as:
3. Graph Weights:
4. Grid Cells
and Object Handover
Images divided into
cells (for speed and constraints):
v Cells have overlap
v Bipartite matching done for
each cell
v Tracks crossing cells are
linked together
VI. Multiple Cameras
v Different Camera Response
Functions (CRFs):
v Inter-camera equalization
using Gamma Function:
v Alpha-blending:
VII. Ground Truth
[Download
sequences information & ground truth here]
VIII. Qualitative Results
v Comparison against
nearest-neighbor greedy bipartite (without constraints):
Track Completeness Factor Track Fragmentation
IX. Qualitative Results
[Download video
for Sequence 1 here]
X. Publications
Vladimir
Reilly, Haroon Idrees, Mubarak Shah,
“Detection
and Tracking of Large Number of Vehicles in Wide Area Surveillance”,
European
Conference on Computer Vision, 2010.
[Download
paper here][High
Resolution]