Hey everybody!
  As finals week draw to a close it looks like the forestry interns will have the time to get some much-needed field work done. Although it's coming close to the wire, or departure date, we're confident we can get our classifications and digitizing done within the next week. Here's little summary of what our contribution to the EcoSynth team has been over the past semester: Digitizing 25x25 field mapping done in previous semesters (February/March), Assisted with several Hexacopter flights (throughout), Surveyed remaining portion of the Knoll (April-present). The work that still needs to be accomplished includes completion of the survey in the Knoll by getting species and crown height measurements, and potentially creating a campus tree-identification guide. Additional related field work completed outside of the internship, and within the environmental mapping class( GES 485) is a complete, and digitized 25x75 meter survey subset into 5 meter grid cells, this survey included the positioning of every tree, large detritus items and stream profile. All of this collected data can hopefully be used to further calibrate the computer vision system, and can be used as a base for further research.

Extra Links: Using backlighting only from an IPhone, Grant Schindler developed this low budget (point cloud it seems) 3D scanning software. While it is nowhere near research grade equipment, He seems to have chosen using a non-natural lighting source to assist in 3d visualization, and for the Iphone it makes perfect sense to use the backlight adjacent to the front-facing camera.

http://www.newscientist.com/blogs/onepercent/2011/04/scan-your-face-in-3d-with-your.html
Grant's page: http://www.cc.gatech.edu/~phlosoft/

This past Saturday, our Environmental Mapping class (GES485) completed an extremely high detail forestry map. The plot was 3, 25x25 meter cells ranging from the top of the hill within the forested area in Herbert run, down to the stream, and back up the opposite slope. The purpose of this survey was to provide a framework for the hypothesis'
of students within 485, and as a longer term source of reference for collected EcoSynth data. We classified as much as was feasibly possible within a day; Fallen trees, stream borders, and any standing tree 2m tall and 1cm wide. As noted in prior blogging, the last survey of this part of campus covered more area, but with less detail. Because we are not exactly sure what the recognizable DBH (diameter @ breast height) threshold of the current camera setup is, this data should become rather useful for calibration, and error checking in the future. One final note regarding the picture at left, each dot is proportional to its relative area the tree takes up on the map, many of them that are very thin trees,
 at this low a resolution, cannot be seen here.

Side question/ attempt at humor of the day:

 If we can get Ecosynth software to recognize objects by color based on previously taken photos, are we developing hue-ristics?

This month's issue of National Geographic on page 30, has a rather interesting
article with many similarities to our current research endeavors. While this
article's research platform is a LiDAR scanning system, the method to gain a
concept of carbon storage within a canopied system, by way of 3D visualization
is fairly similar. The author notes the large scale tree planting programs currently
 underway in China and their effect on the reduction of CO2 emissions. Long running
political debates regarding carbon trading are summarized in this online abridged
 version of the full article, but the focus is following Greg Asner's team
of ecologists and their "CAO" program.  The teams involvement in the REDD
(Reduce Emissions from Deforestation and Degradation) have boosted confidence in
providing a cost-effective and high precision method in estimating carbon stores
 within the Amazon. Results found that older, more developed forests stored as much
 as three times the amount of carbon than those composed primarily of secondary growth.
One abstract application of their airborne observatory noted using this 3D imaging
 platform to map termite mounds in savannahs, I suppose this would have some application
 for unmanned vehicles, because hexacopters are resistant to termites and fire ants….

Links:

http://ngm.nationalgeographic.com/2011/05/big-idea/tree-map-text 

http://cao.ciw.edu/

     After an afternoon with Mariah, we both realized we were over-thinking the compass with a lack of confidence, after believing we were working off the wrong  bearing (and briefly losing some equipment), begging Jonathan for help we went back out in the field only to find we weren’t far off in the first place. So today we’ll be making some effective plotting from 2 to at least 4 pm. Following the laying out of flags, each plot will have the DBH and species identified, and following the physical classification this should be digitized. So as of yet, 23 plots to go, and we’ll see how today goes.
    The 485 class was not so lucky this past weekend, getting rained out on Saturday for our soil profile collections, this ended up pushing our planned field days for our individual group projects back as well. Some of these projects have some immediate impact on the ecosynth project in part, Chris Leeney, Lauren Colburn, Nathan Rolls and I have chosen our final projects in furthering accuracy of Chris’ groundsynth work. Nathan is interested in tree growth and density characteristics from streambeds, and in this case Herbert Run makes for a good study area. Laurens work concerns the degree to which Groundsynthing can recognize trees (what is the minimum recognizable DBH). My work will be in depicting the covered area in 5x5 grid cells that are manually drawn, including high levels of detail that can be noticed in Groundsynthing to further spatial referencing in small studied areas, and for recognition of erroneous data ( the maps will include any tree above 2m height, fallen trees, stumps, etc.) Hopefully the undergraduates field day doesn’t get pushed back by seasonal weather patterns once again.

With the weather picking up and tree species becoming recognizable once again, we have the opportunity to finish surveying the Knoll. Using a standardized procedure Mariah and I will be going to each of the 25 by 25 cells ( 23 total) and taking a number of measurements. Identification of the species within, finding their diameters at breast height and taking crown height measurements using our trusty laser hypsometer of the 5 largest trees within the plot will be the general order to follow. Because our data will be correlated with over- canopy imagery, collecting the crown heights of smaller trees that will only be covered by the canopies of larger trees are not totally necessary. Our study area has come under significant change since last summer. Renovation and management of the stream nearest to the parking garage was done from January till just last week (major construction ceased last week at least) removing around 250 square meters of vegetation along the run. Next Wednesday afternoon we will begin the survey and should be complete within two or three weeks.

Within the past few weeks digitizing the field collections was completed, this was accomplished by taking Jonathan and Evan's collected field data and creating a referenced shape file over the area of Herbert Run. The work was split by Mariah and I, in the picture above her points are green, mine are yellow. The data was in the form of hand-drawn grids and estimated point positions within the subplot were numbered including the species and estimated DBH of the described tree.

    The method I used in creating this representative tree population distribution is fairly straight forward. Each subplot drawing was overlaid with a transparency in which I attempted to equally partition the cell into 25, 5x5 meter subcells. A 5x5 meter subgrid polygon file was supplied for the Herbert run area and the drawn points were transferred.

 One of the main issues with the supplied data that may cause minor error was the case of trees that split at the trunk. These were denoted on the drawn grids as two dots, and frequently interpreted as two separate trees in very close proximity, rather than the same tree with two DBH's. This data can be fixed but its overall error effect on canopy data may be negligible.

  It's useful to note again that these points are NOT the true locations of the trees, but this set gives a representation of where the species occur. The next major step of this dataset will be collecting the heights of the given trees with the laser hypsometer, or collecting the data in 5 x 5 meter cells or both....we'll see which technique proves fit in the coming weeks!