Performance Analysis of Deep Learning Based Object Detection Algorithms for Satellite Imagery

Abstract

Object detection is a vital step in satellite imagery-based computer vision applications such as precision agriculture, urban planning and defense applications. In satellite imagery, object detection is a very complicated task due to various reasons including low pixel resolution of objects and detection of small objects in the large scale (a single satellite image taken by DigitalGlobe comprises over 240 million pixels) satellite images. Object detection in satellite images has many challenges like class variations, multiple object poses, high variance in object size, illumination and dense background. This study aims to compare the performance of existing deep learning algorithms. We created the dataset of satellite imagery to perform object detection using Convolutional Neural Network based frameworks like Faster RCNN, YOLOv3, SSD and SIMRDWN. In addition to that, we also performed an analysis of these approaches in terms of accuracy and speed using the developed dataset of satellite imagery. The results showed that SIMRDWN has good accuracy of 97% and speed on high-resolution images, while Faster RCNN has good accuracy of 95.31% on the standard resolution (1000 × 600). YOLOv3 has good speed and accuracy of 94.20% on standard resolution (416 × 416) while on the other hand SSD has just good speed on standard resolution (300 × 300) with accuracy of 84.61%.