Fall 2003 Capstone Project: Seismic Sled Data Noise Filtration

This project is concerned with taking seismic data from the geology department and figuring out how to remove noise from the data to get a more accurate reading. The land-streamer was created to make recording seismic data easier. The land-streamers are wooden sleds with one geophone attached to each of them. These sleds are strung together and drug behind a 4X4 vehicle and pulled down the line of interest. The sleds are stopped at specified increments and then a seismic blast is created causing vibration in the ground for the geophones to pick up. Then they are moved to the next location. When seismic data is recorded by the land-streamer, due to the vibration in the wood and other factors, unwanted noise enters the final data set.

The goal of this project is to first convert the data from the SEG-Y format into a MATLAB readable format. Next we take this data and using the Fourier wave transform we analyze the wave and remove a large amount of the noise. Finally, we compare the adjusted data set to the already known data set to see how close we were able to match the two readouts. This will help to create a more efficient way for seismic data to be recorded.

Unfiltered seismic data
This is a picture of the unfiltered data using discrete Fourier transform. Notice that there is continuous distribution of frequencies, some frequencies are noisy, others are signal.
Filtered seismic data
This is a picture of the frequency filtered data using discrete Fourier transform. Notice that all non-peak frequencies have been removed from the data.