Tracking Migratory Birds Around Large Structures

1
Tracking Migratory BirdsAround Large
Structures Presented by Arik Brooks and
Nicholas Patrick Advisors Dr. Huggins, Dr.
Schertz, and Dr. Stewart Senior Design Project
2003-2004Bradley UniversityDepartment of
Electrical and Computer Engineering
2
Project Background

• Every year, many birds are killed when their
  migration path takes them near tall structures on
  overcast nights.
• One widely accepted theory on why this happens is
  that the birds do not want to leave the lighted
  area near a structure and end up running into it.
• Wildlife biologists would like to study this
  phenomenon.

3
Outline

• Project summary
• Previous Work
• Detailed description
• System block diagram
• Subsystems
• Results

4
Outline

• Test Plan
• Datasheet
• Conclusions
• Suggestions for future work
• Questions

5
Project Summary

• The purpose of this project is to implement a
  system to track the flight paths of birds in
  real-time via stereoscopic imaging.
• The desired system output is a display depicting
  a 3-D representation of the trajectories of the
  birds, and data relating to the trajectories.

6
Previous Work

• 2003 seniors Brian Crombie and Matt Zivney
• Results basic object position location in a
  laboratory environment with major limitations.
• The groundwork laid out in their project
  (algorithms, design equations, software
  organization, etc.) was used as a starting point
  for our system.

7
Detailed Description
8
System Block Diagram
System
9
Hardware Block Diagram
10
Subsystems

• Cameras
• Frame Grabber
• PCs/Network

11
Camera Subsystem

• Includes two cameras mounted in parallel a known
  distance apart allowing objects to be located in
  space.
• Inputs
• Photons -- Images collected by the cameras
• Synchronization -- Internal line lock
• Outputs
• Data -- Image data transmitted to the frame
  grabber
• Operation in System
• The cameras capture images at a rate dictated by
  the speed of the preprocessing algorithm

12
Frame Grabber Subsystem

• The frame grabber simultaneously captures images
  from both cameras and supplies the digitized
  image data to the PC.
• Inputs
• Data -- Image data (NTSC format) from the cameras
• Setup -- Information from the PC
• Outputs
• Image Data to PC
• Operation in System
• The frame grabber operates at a rate dictated by
  the speed of the preprocessing algorithm

13
PCs/Network Subsystem

• Two PCs are networked together to divide
  computation between the preprocessing and
  trajectory calculation computers.
• Inputs
• Image Data -- Arrays of intensity information
• Calibration Input -- Calibration data for the
  cameras being used
• Outputs
• Display GUI showing trajectories plotted in a
  three dimensional representation
• Statistics -- Pertinent data calculated from bird
  trajectories
• Raw Data -- Data file containing all preprocessed
  data
• Operation in System
• The PCs and network operate continuously

14
Results
15
Preprocessing Software
16
Streamlined Preprocessing in C

• Implement faster centroid location code.
• Perimeter search vs. pixel-by-pixel search
• Improve background subtraction algorithm
• Fixed number of frames averaged for background
  to 256
• Current frame added using shift operations
  instead of multiplies/divides
• Stored background is 16 bits
• upper 8 bits are image data
• lower 8 bits for accumulating round-off error

17
Streamlined Preprocessing in C

• Improve background subtraction algorithm
• Speed Improvements (640x480, threshold image, do
  not find objects)
• Old -- 10.6 Frames per Second
• New 15.9 Frames per Second
• Updating average every 60 frames
• Without find object function -- 24 Frames per
  Second
• With find object function -- 18 Frames per Second

18
Preprocessing in C
19
Trajectory Determination in MATLAB

• Code to correlate objects between 2 cameras and
  over time restructured
• Added predictive searching to significantly
  improve tracking ability
• Improved graphing techniques real-time operation
• Implemented GUI for easy user interface

20
Trajectory Determination comparison - two tennis
balls swinging

• System from last year

21
Trajectory Determination comparison - two tennis
balls swinging

• Current algorithm

22
Trajectory Determination Software
23
Trajectory Determination in MATLAB

• Predictive Search Method
• Search for a new point within a sphere defined
  by
• Center at the location (x,y,z) predicted by the
  previous two points in the trajectory and the
  time taken between frame-grabs
• Radius determined by average bird velocity, time
  between frames, current velocity, and distance
  from the cameras

24
Trajectory Software GUI
25
Test Plan

• There will be three primary test procedures that
  will be performed to verify the system
  specifications
• Location Accuracy
• Max/Min Distance from Cameras
• Max Objects

Tennis Ball dispenser used in accuracy testing
26
Test Plan

• Location Accuracy
• Capture data at known heights using a
  stationary object and balls dropped from the
  tennis ball dispenser. Compare theoretical to
  experimental.
• Max/Min Distance from Cameras
• Repeat Location Accuracy experiment at extremes
  of range.
• Max Objects
• Nerf Guns!!!

27
T-Bird Accuracy Test 1
28
T-Bird Accuracy Test 2
29
T-Bird Demonstration
30
T-Bird Demonstration
31
Datasheet

• Average Migratory Bird Diameter 0.152 m
• Average Migratory Bird Speed 8.9409 m/s
• Max of Objects Tracked Simultaneously TBD
• Max Distance from Cameras 20 m
• Min Distance from Cameras 3 m
• Max Location Error (theoretical) 0.375 m
• Light Level Sensitivity
• Lab Cameras 0.22 Lux
• Low Light Cameras 0.0002 Lux
• Max Framerate 15 FPS
• Total Volume of Space Observed 606 m3
• Separation of Cameras assumed for calculations
  0.5 m

32
Conclusions

• Real-time tracking of multiple objects was
  achieved in a laboratory setting.

33
Suggestions for Future Work

• Implement boom (mechanical system and controls)
• Obtain and integrate high end cameras
• Optimize code (analyze algorithms, streamline
  processes)
• Port MATLAB to C
• Investigate feature detection methods for
  improved target recognition

34
Tracking Migratory BirdsAround Large
Structures Questions?