Object detection from satellite imagery

• Main Author: Seah, Yi Xuan
• Other Authors: Lu Shijian
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2020
• Subjects: Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence; Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision
• Online Access: https://hdl.handle.net/10356/138233
• Institution: Nanyang Technological University

Satellite imagery has been used to observe and collect information about the earth for decades. Objects such as vehicles, planes, and ships can be detected from these imageries. However, as the imageries often lack contrast details that are critical to the effectiveness of fast and accurate detection techniques. Thus, Machine Learning techniques such as Deep Learning are required to process the imageries in a fast and accurate manner. This report will be investigating Deep Learning-based object detection techniques. Image classification techniques such as VGG and ResNet will be studied to determine which is more suitable for satellite imagery. Object Detection techniques such as R-CNN, Fast R-CNN, and Faster R-CNN will be also be studied to understand the progress of object detection methods. Lastly, tests using metrics such as mean average precision (mAP) and inference time will be used to determine the suitability of object detection for satellite imagery.