Weighted Ensemble Object Detection with Optimized Coefficients for Remote Sensing Images

Atakan Körez,Uçman Ergün,Necaattin Barışçı,Aydın Çetin

doi:10.3390/ijgi9060370

Atakan Körez, Uçman Ergün + Show 2 more

Open Access

https://doi.org/10.3390/ijgi9060370

Copy DOI

Abstract

The detection of objects in very high-resolution (VHR) remote sensing images has become increasingly popular with the enhancement of remote sensing technologies. High-resolution images from aircrafts or satellites contain highly detailed and mixed backgrounds that decrease the success of object detection in remote sensing images. In this study, a model that performs weighted ensemble object detection using optimized coefficients is proposed. This model uses the outputs of three different object detection models trained on the same dataset. The model’s structure takes two or more object detection methods as its input and provides an output with an optimized coefficient-weighted ensemble. The Northwestern Polytechnical University Very High Resolution 10 (NWPU-VHR10) and Remote Sensing Object Detection (RSOD) datasets were used to measure the object detection success of the proposed model. Our experiments reveal that the proposed model improved the Mean Average Precision (mAP) performance by 0.78%–16.5% compared to stand-alone models and presents better mean average precision than other state-of-the-art methods (3.55% higher on the NWPU-VHR-10 dataset and 1.49% higher when using the RSOD dataset).

Highlights

IntroductionWith the enhancement of remote sensing technologies, remote sensing image object detection has gained popularity thanks to its successful civil and military applications (e.g., urban monitoring, traffic monitoring, agricultural applications, and landscape planning)
At present, with the enhancement of remote sensing technologies, remote sensing image object detection has gained popularity thanks to its successful civil and military applications
An experimental setup environment was created to evaluate the performance of the proposed model on the NWU-VHR10 and Remote Sensing Object Detection (RSOD) datasets

Summary

Introduction

With the enhancement of remote sensing technologies, remote sensing image object detection has gained popularity thanks to its successful civil and military applications (e.g., urban monitoring, traffic monitoring, agricultural applications, and landscape planning). Many studies on the detection of objects in remote sensing images are available in this field of study. Peicheng et al [2] proposed a novel and effective approach to train a Rotation-Invariant Convolutional Neural Network (RICNN) model for advancing the performance of object detection, and Wang et al [3] used a skip-connected encoder–decoder model to extract multiscale features from a full-size image. Wu et al [4] detected remote sensing objects using Fourier-based rotation-invariant feature boosting (FRIFB). Cheng et al [7] proposed a multiclass object detection feedback network (MODFN) using a top-down feedback mechanism based on a traditional feedforward network. Cheng et al [10] developed a rotation-invariant framework based on the Collection of Part Detectors (COPD) for multiclass object detection. For a comprehensive and recent survey, see [17]

Objectives

Methods

Results

Discussion

Conclusion