Chapter 14 - Object detection in remote sensing

Abstract

Situated between scene-level classification and pixel-level semantic segmentation, object detection represents one of the most fundamental tasks in computer vision and remote sensing. It enables the automatic localization and recognition of distinct objects, such as wildlife, aircraft, ships, or vehicles, across a broad range of sensing modalities, including optical RGB, multispectral, hyperspectral, LiDAR, and SAR imagery, and from platforms spanning UAVs to satellites. This chapter introduces the principles and methodologies of object detection in the context of Earth observation, tracing its evolution from early handcrafted approaches to modern deep learning-based techniques. It outlines the main stages of the detection pipeline, highlights seminal models that shaped the field, and discusses an illustrative application of aircraft classification in SAR imagery. The chapter concludes with practical guidelines, open challenges, and future directions for advancing object detection in remote sensing.