Yesterday, Chris, Jeff and I took Vernon on a mission to survey a brand new area.
In the afternoon, we continued working on our own research. Jeff and I are working together on the identification of potential heritage sites (shipwrecks, planes etc.) from side scan sonar images. Our method consists of two stages, object proposal and classification. Our object proposal pipeline uses rectangular filters to look for shadows of objects in the sonar scans. It works quite well on our test data set and is able to capture all the targets.
 |
| Vernon on the industrial boat |
In the afternoon, we continued working on our own research. Jeff and I are working together on the identification of potential heritage sites (shipwrecks, planes etc.) from side scan sonar images. Our method consists of two stages, object proposal and classification. Our object proposal pipeline uses rectangular filters to look for shadows of objects in the sonar scans. It works quite well on our test data set and is able to capture all the targets.
However, the purpose of object proposal is to filter out most uninteresting areas in the scans in a quick and simple fashion, so it produces a lot of false detections. We are working on the following classification stage that uses machine learning techniques to reduce false alarm rate and make our detections more precise.
The rest of the team had a productive day working as well. Roslyn is working on getting some new models ready for the web, Sam and Amy worked on getting openCV up and running with their openGL application and Jane worked on literature review of PRM and implemented an RRT in python.
