Point Cloud Feature Engineering for Undersea Pipelines

Abstract

Sonar based point cloud data is acquired by remotely operated vessels during undersea pipeline inspections. Features extracted from this point cloud data are used in point cloud classification tasks in production, hence making feature engineering an important part of the whole point cloud processing workflow. Finding the suitable features to be used in point cloud classification is a challenging task. Architecting the right infrastructure for the process could also contribute to this complexity. In this talk we go through the most important features, extracted from the undersea point clouds via intelligent algorithms including machine learning and how we tackle the challenges of running this task in a commercial world project.

Bio

Amir Rastar is a senior innovation engineer at Fugro working on point cloud classification and image processing of the undersea pipeline surveys. He is a member of the compute backend team (Mantis Shrimp) for Fugro’s Sense Pipelines project which develops, deploys and maintains point cloud processing services and serverless cloud infrastructure. Before joining Fugro, he was a data scientist at Aeroqual, an air quality sensor manufacturing company based in Auckland, NZ. During that time, he analysed and developed predictive models for air quality sensor data.