MIT Lincoln Laboratory MIT Student Technical Assistant-x-terms Program--Spring 2019 in Lexington, Massachusetts

Requisition ID: [[id]]

The Advanced Capabilities and Systems Group provides assessments of novel technologies and system concepts to meet significant and pressing defense and intelligence needs and, where appropriate, rapidly develops prototype solutions to demonstrate concepts or provide fieldable capability. To accomplish these goals, the group taps Laboratory-wide expertise and couples this with the group’s strong systems analysis and prototyping capabilities. Modeling, often supported by quick measurements and tests, is used to evaluate the feasibility of proposed solutions to problems, as well as to creatively develop new alternatives. Products of this assessment process include rapid prototyping efforts as well as briefings and proposals for follow-on development efforts. Where a rapid capability is sought, the group often leads multigroup coalitions in the execution of these efforts.

Project Focus: Littoral Geolocation Estimation-- the goal of this project is to assess the capability of geo-locating a littoral image based on the derived shape of the shoreline.

Previous group work has established the uniqueness of the shape of a finite segment of a shoreline assuming perfect knowledge of the terrain. We showed how unique, over how large a coastal universe, and how robust the uniqueness is when subject to rotation and scaling distortions. More recent work has analyzed the ability to classify digital imagery at the pixel level by using a trained Convolutional Neural Network. We developed the processing chain to pull imagery and truth from databases, and classify against a trained CNN (TensorFlow software) that we trained. We compared the pixel-level accuracy results from training on imagery from the same or different parts of the world.

For this project, the goal is to combine the results from the previous efforts to build a nominal capability of littoral geolocation from actual imagery. The candidate will assess the resulting capability for accuracy and robustness to various algorithmic inputs. If time allows, the candidate will also assess other key issues such as the impact of the quality of the truth database.

The candidate should be pursuing an undergraduate degree at MIT and have proficiency with Matlab, and familiarity with CNNs (preferably Tensorflow or Matlab’s neural network toolbox), to quickly leverage, adapt and optimize the code bases. Demonstrated analysis skills are a plus.

MIT Lincoln Laboratory is an Equal Employment Opportunity (EEO) employer. All qualified applicants will receive consideration for employment and will not be discriminated against on the basis of race, color, religion, sex, sexual orientation, gender identity, national origin, age, veteran status, disability status, or genetic information; U.S. citizenship is required.