MIT Lincoln Laboratory

MIT Lincoln Laboratory is a United States Department of Defense federally funded research and development center chartered to apply advanced technology to problems of national security.

Group 91 Co-op

From July to December 2022, I had the privilege of working at MIT Lincoln Laboratory in Group 91. During this time, I contributed to two significant projects: SaMSoN and another project, the details of which are confidential.

Projects

SaMSoN

The purpose of SaMSoN is to analyze vegetation changes in satellite imagery related to hurricane activity. When a hurricane strikes, direct measurements aren’t feasible, as deploying instruments and gathering extensive data isn’t practical. We know hurricanes can cause beach erosion and accretion, but to what extent? And what role does dune restoration play—does it help prevent significant erosion during hurricanes? How effective is dune restoration overall? SaMSoN aims to answer these questions and shed light on the impact of hurricanes on coastal vegetation and dune resilience.

SaMSoN Summary

What I Contributed:

I developed a program based on CoastSAT to collect remote sensing data of shorelines, enabling the tracking of changes over time. This involved utilizing Python scripting to analyze satellite imagery and contribute to the field of environmental monitoring. The project aimed to assess coastal changes, demonstrating my proficiency in coding and satellite image analysis for environmental science applications. Additionally, I presented the outcomes of this work at the International Conference on Coastal Engineering (ICCE 2022), highlighting my ability to translate technical achievements into impactful presentations for a broader audience.

Challenges:

We want to identify changes in vegetation in satellite imagery.
We want to detect shorelines that are sand water boundaries and vegetation water boundaries accurately in satellite images.

Solutions Implemented:

Trained a machine learning classifier for vegetation identification.
Developed a new shoreline detection algorithm.
Implemented image analysis techniques for precise shoreline delineation.

Lessons Learned:

Gained proficiency in machine learning techniques for image classification.
Enhanced my understanding of satellite imagery analysis.
Gained experience in problem-solving and innovation.

Presenting my work from CURES and MITLL at ICCE 2022

CoastSAT Program Overview

My Program Overview with changes highlighted in red

The images above show the original code and what I changed in the code. I added in a new classifier to detect vegetation and a new algorithm using Canny Edge Detection and K-Means Clustering to detect shorelines.

CoastSAT Final Output

SaMSoN final output. For the NDVI Image, green signifies a high likelihood of vegetation, while blue represents other elements

Notably, my code includes vegetation data as well as all the normal data CoastSAT collects. Furthermore, it incorporates an NDVI (Normalized Difference Vegetation Index) view, a method for assessing the likelihood of vegetation presence. This improvement not only enriches the output but also provides a more comprehensive and visually intuitive understanding of the data.

The left shows CoastSAT vs my code, and the right shows a shoreline study of the same area. The left graph had an R^2 value of .98 between my code and CoastSAT. While lacking specific comparative data of the left and right images, a visual inspection reveals striking similarities in the shapes of the graphs.

Analyzing the shoreline transects over time revealed a capability in detecting significant environmental events, notably a hurricane occurrence. Around September 2019, a distinctive downward trend in the data emerged, persisting for the remainder of the year. The graph shows a substantial shift in shoreline location relative to the mean during this period, coinciding with the occurrence of Hurricane Dorian. Manual inspection of the satellite imagery further corroborates this observation, showcasing a discernible change in the shoreline before and after the hurricane.

Left image shows satellite image of study area. Middle graphs show shoreline location over time. Right images show study location before and after hurricane.

The left image encapsulates an analysis of vegetation pixels, with the blue line representing the initial input data. Notably, in an area undergoing dune restorations, the long-term trend of vegetation pixels should exhibit an upward trajectory. This positive trend is illustrated by the red line in the graph. Furthermore, the graph unveils an observation of seasonal variations, as depicted by the yellow line. It discerns the annual cycle, showcasing heightened vegetation during the summer months and reduced levels during the winter.

Second Project (Signed NDA)

I undertook the design and prototyping of attachment parts for real-world objects using FDM 3D printing. I also calibrated cameras using both checkerboard and charuco boards. I implemented tracking methods using aruko markers scattered around a specific area, capturing the position (x, y, z) in the real world, as well as monitoring the object's yaw, pitch, and roll. Additionally, my role extended to collaborating on the development of a measurement system, where I worked hands-on with a magnetometer. This comprehensive project showcased my diverse skill set, encompassing design, prototyping, calibration, 3D modeling, and system integration.

Page updated

Google Sites

Report abuse