Introduction To Robotics Lab 2

Zachary Greenberg

Data Collection Process

  1. Use a tape measure to measure the required distance from the board. Place a masking tape marker at each distance
  2. Find the center of the board. Place a tape marker at each site perpendicular to the last, marking the center of the board at each distance
  3. Align the camera to the intersection point of the first tape marker. In this case a ridge on the back of the iphone case was used to center the camera. The bottom of the iphone should be placed on the floor so that the height is always the same.
  4. Repeat for the other distances.
  5. Once all distances for one board are measured, repeat steps 3 and 4 taking care to align the center of the new board with the center of the old one.

Results

Test Distance Calculated (in) Original Filtered Threshhold Segments Ball Identification
Small Board 2 Feet 22.7875 Small Board from 2 Feet Small Board from 2 Feet Small Board from 2 Feet Small Board from 2 Feet Small Board from 2 Feet
Small Board 4 Feet 46.7103 Small Board from 4 Feet Small Board from 4 Feet Small Board from 4 Feet Small Board from 4 Feet Small Board from 4 Feet
Small Board 8 Feet 93.6504 Small Board from 8 Feet Small Board from 8 Feet Small Board from 8 Feet Small Board from 8 Feet Small Board from 8 Feet
Small Board 16 Feet 183.4399 Small Board from 16 Feet Small Board from 16 Feet Small Board from 16 Feet Small Board from 16 Feet Small Board from 16 Feet
Large Board 2 Feet 21.7090 Large Board from 2 Feet Large Board from 2 Feet Large Board from 2 Feet Large Board from 2 Feet Large Board from 2 Feet
Large Board 4 Feet 45.3609 Large Board from 4 Feet Large Board from 4 Feet Large Board from 4 Feet Large Board from 4 Feet Large Board from 4 Feet
Large Board 8 Feet 91.2317 Large Board from 8 Feet Large Board from 8 Feet Large Board from 8 Feet Large Board from 8 Feet Large Board from 8 Feet
Large Board 16 Feet 184.8334 Large Board from 16 Feet Large Board from 16 Feet Large Board from 16 Feet Large Board from 16 Feet Large Board from 16 Feet


Trial Abosolute Error Percent Error
Small Board Small Absolute Error Small Percent Error
Big Board Big Absolute Error Big Percent Error

Analysis

In general the absolute estimation error increased as the picutre got farther away. The tennis balls in farther away images were smaller and thus harder to detect. This also meant that filtering had a greater effect on them. This made it harder to distinguish tennis balls from noise. However the percent error gave a less clear trend. On the large board, as the distance increased exponentionally error only increased linearly, so the percent error decreased. However for the small board the 16 feet away picture had a large spike in percent error. One reason for this could be that the in that picture, after filtering the top 2 tennis balls lost a lot of mass. This led to an inacurate center of mass, which makes the distance calculations off. In the large board, all the balls were filtered evenly so this error didn't occur. The worst image in terms of percent error was the large board from 2 feet. This is becuase the board was so big, it almost didn't fit in the frame. Either the camera had to be tilted to capture the entire image, or parts were cut off. Both tilt and occlusion of the tennis balls leads to bad error. However the error was still less than 10 percent which is quite good.