Technical Report Number
Geophysical phenomena are often three-dimensional, time-variant, and physically large, making them difficult to measure. As wireless sensor nodes become cheaper, smaller, and more powerful, using a sensor swarm as a sampling framework seems to be a viable approach to this problem. However, samples from the network can be sparse and unstructured, creating an incomplete picture. Thus, the question addressed by this project is the following: How can we easily construct and interpret the best model of the underlying reality over some domain, given a possibly sparse and irregular set of samples? We present the design, implementation, and evaluation of the Signal Reconstruction Panel, a graphical program for tackling this problem. Using a method of Support Vector Regression, we demonstrate the program's performance on a variety of data sets including the Collisionless Terrella Experiment (CTX), the Active Magnetosphere and Planetary Electrodynamics Response Experiment (AMPERE), the Poker Flats Incoherent Scatter Radar (PFISR), and GreenCube5, a detailed study of the Ompompanoosuc River flow.
Dartmouth Digital Commons Citation
Guinther, Jonathan H., "Interpreting and Reconstructing Data from Sensor Networks" (2013). Computer Science Technical Report TR2013-735. https://digitalcommons.dartmouth.edu/cs_tr/360