A Study on Applying Learning Techniques to Remote Sensing Data
Abstract
A major issue with data-hungry deep learning algorithms is the lack of annotated ground truth for specific applications. In this thesis, we explore the challenges of applying artificial intelligence techniques for remote sensing data which lacks the availability of large annotated datasets for training in comparison to regular imagery data. We first tackle the problem of improving object tracking in Wide Area Motion Imagery data by using a semantic segmentation model to detect false tracker points on buildings. The combination of image understanding techniques with tracking has the potential to improve track quality and aid automatic situation assessment. We use a manually annotated dataset which confined us to work with a small dataset. We propose a solution for this problem by developing a framework for automated annotation of remote sensing data. We pick satellite imagery due to the high volume of Earth Observation Satellites available today, coupled with crowd-sourced map data that can enable a new means for automated annotation of remote sensing data for applications previously constrained by the lack of available datasets. In the second part of this thesis, we present an automated pipeline for collecting and labeling satellite imagery to facilitate building custom deep learning models. We demonstrate this approach by automatically collecting labeled imagery of solar power plants and building a classifier to detect the presence of such structures. This framework can be used to collect labeled satellite imagery of any object mapped by spatial databases.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 04, 2020
- Accession Number
- AD1097520
Entities
People
- Aswathnarayan Radhakrishnan
Organizations
- Ohio State University