Last week we had our official field day to collect all our data for the 6 chosen methods, all in all meaning a productive 12 hours in the stream. The day began with a simple task of placing our thirty pre-made and marked control points. From there we then began to measure distances between the control points in two teams of two and then compared our results which surprisingly enough were spot on for the majority of them. Next we each began photo methods. We used the GPS feature of the camera to get longitude and latitude coordinates of the waters edge on the banks. We used a tripod stand for the “twinning” cameras method to take pictures that could be used to create 3D images of the stream. As with the first practice “field day” the square method was also utilized but with slight modifications this time. In addition to using the square to obtain horizontal and vertical lines, we also placed the square in three different locations and tried to include as many control points in pictures as possible as a means of testing the accuracy of the perspective correction. David and Kyle began the photosynth method which consisted of walking up and down the center of the bank, and up and down the center again first looking at one bank and then the other taking pictures all the while (over 500 pictures between the two of them for this method alone). Next was the video method wherein 15 second videos were taken while walking down either bank. The non-photo methods were pretty frustrating but they turned out a lot better than expected once we got a system down and started. Maybe we didn’t think them through as well as the others or expected the control points to be on the edge, but either way, we found a solution quickly.
Here are some field day pictures. For many more pictures and a time lapse video: http://streamoutlines.blogspot.com/
Since the major field day, we have spent all of our time analyzing data. The GPS method consisted of using an equation to convert degrees of longitude and latitude to x, y coordinates that could be plotted on the graph. This method had by far the highest percent error (over 200%) and therefore it was determined that the camera GPS was not precise enough. For the non-photo methods again, x, y coordinates were established and the angle was determined to graph the stream outline, both showing less than 3% error. The twinning cameras/3D/video we have not analyzed yet due to time constraints and focusing a lot of our time on what we believe to be the most effective method and certainly the most novel method of them all, the square method. In these last couple of days we have manually picked out the outline, the control points, and the square corners by hand using GIMP on hundreds of images that are part of the square method to upload eventually into a Matlab program that will automatically correct for radial distortion and perspective allowing us to create a map and calculate control point error for this method. But before these images are able to be loaded into Matlab to correct for perspective and barrel distortion they are run through the StreamGUI. This program allows users to load the images and click and label the corners of the square and the control points as well as setting the square’s side length and the horizontal field of view. We hope the percent error for this method will be the lowest of them all as it seems to be continually lowering as more and more pictures are completed. We are excited to share our results with everyone tomorrow!!