Ongoing thesis. Analyzing sonar data. Student: Pyojeong Kim Advisors: Arijit Das, Sean Kragelund, Glenn Cook.
UUV & Mine Warfare (mission): Image Credit: http://www.bluebird-electric.net @arijit_da
Sonar Data File Each mission generates about GIG to a Terabyte of data. Depends on the duration. Over decades, Terabytes of files have been generated. Files need to be catalogued (using automated techniques). @arijit_da
Sonar Data File Each binary file has 2 parts. Metadata file name, timestamp, latitude, longitude, depth and more. Image Extraction and processing (future thesis topic) @arijit_da
Sonar Image Processing. Image Credit: https://tinyurl.com/y62xpu2g @arijit_da
Cataloging: Classic Hadoop problem. 100s of thousands of files. Same processing on each file. Map Phase is enough to collect metadata. @arijit_da
MapReduce. @arijit_da
Non Hadoop processing. File format provided by the sonar vendor. Use Java typecasting to map the raw bytes to the fields. For large number of files: Time consuming on single laptop, desktop, server. @arijit_da
Oracle DB Table. @arijit_da
Oracle Table Data. @arijit_da
HIVE Table. @arijit_da
HIVE Table Data. @arijit_da
Sqoop: HDFS HIVE back to Oracle DB Image Credit: https://www.dezyre.com/ @arijit_da
Oracle DB &APEX. @arijit_da
Hadoop in a box. ImageCredit: https://www.casesbypelican.com/ @arijit_da
AI to predict the Depth. Multilayer perceptron (MLP) regression. Other AI Techniques. @arijit_da