Effective Airplane Detection in High Resolution Satellite Images using YOLOv3 Model | IEEE Conference Publication | IEEE Xplore

Effective Airplane Detection in High Resolution Satellite Images using YOLOv3 Model


Abstract:

Airplane detection is an artificial intelligence problem that can be solved either by the classical machine learning methods or by the cutting-edge deep learning processe...Show More

Abstract:

Airplane detection is an artificial intelligence problem that can be solved either by the classical machine learning methods or by the cutting-edge deep learning processes. Generally, object detection is a highly challenging task as it requires recognizing the desired objects in the image in addition to localizing their positions by drawing a rectangular box with a confidence percentage. To obtain a high level of detection accuracy, the developed models are to be trained with hundreds or even thousands of satellite images; however, these images are limited, with low spatial resolution and high expenses. In recent days, deep learning algorithms have proved to excel in numerous object detection tasks due to the availability of huge amount of data along with the continuous advances in computers' power technology. This study investigates the problem of airplane detection from satellite imagery by proposing a simple, but an effective tool to achieve excellent detection rates. Our proposed method is mainly based on You Only Look Only (YOLOv3) model, a cutting-edge object detection technique, which is faster and more accurate than most of the traditional and well-known approaches such as Region-based Convolutional Neural Network (R-CNN). According to the experimental results, the overall accuracy of airplanes' detection, using KhalifaSat images, is 97.64% with a reasonable computational time. Our findings prove that the proposed YOLOv3 model is robust and efficient in detecting airplanes of different dimensions in high resolution satellite imagery.
Date of Conference: 24-25 November 2021
Date Added to IEEE Xplore: 27 December 2021
ISBN Information:
Conference Location: Dubai, United Arab Emirates

References

References is not available for this document.