Multi-Stage Feature Enhancement Pyramid Network for Detecting Objects in Optical Remote Sensing Images

Abstract

The intelligent detection of objects in remote sensing images has gradually become a research hotspot for experts from various countries, among which optical remote sensing images are considered to be the most important because of the rich feature information, such as the shape, texture and color, that they contain. Optical remote sensing image target detection is an important method for accomplishing tasks, such as land use, urban planning, traffic guidance, military monitoring and maritime rescue. In this paper, a multi stages feature pyramid network, namely the Multi-stage Feature Enhancement Pyramid Network (Multi-stage FEPN), is proposed, which can effectively solve the problems of blurring of small-scale targets and large scale variations of targets detected in optical remote sensing images. The Content-Aware Feature Up-Sampling (CAFUS) and Feature Enhancement Module (FEM) used in the network can perfectly solve the problem of fusion of adjacent-stages feature maps. Compared with several representative frameworks, the Multi-stage FEPN performs better in a range of common detection metrics, such as model accuracy and detection accuracy. The mAP reaches 0.9124, and the top-1 detection accuracy reaches 0.921 on NWPU VHR-10. The results demonstrate that Multi-stage FEPN provides a new solution for the intelligent detection of targets in optical remote sensing images.

A Survey of Small Object Detection Based on Deep Learning in Aerial Images

“…Problem solved Optimization Strategies CF2PN [84] Cross-level information exchange Cross-scale fusion module (CSFM) HawkNet [85] Cross-level information exchange Gated context-aware module (G-CAM) FPN-based [86] Cross-level information exchange Receiver field expansion block ABNet [87] Cross-level information exchange Adaptively combining multiscale features MSGN [88] Cross-level information exchange Backward semantic guided filtering (BSGF) MFEPN [89] Cross-level information exchange CAFUS+FEM SCANet [90] Inadequate expression RFEM+SCFM MVNet [91] Inadequate expression MRBs+MRFEM MSFC-Net [92] Inadequate expression Composite Semantic Feature Fusion (CSFF) SRAF-Net [93] Inadequate expression SE-FPN+SADH VSRNet [94] Inadequate expression VRG+SRG mSODANet [95] Inadequate expression Dilation Convolution for multiple scales SGFTHR [96] Channel information loss Structure Guided Feature Transformation (SGFT) MFAF [97] Channel information loss FI+SAW+CSP M2S [98] Channel information loss Improving feature extraction and feature refinement Info-FPN [99] Feature misalignment Feature alignment module (FAM) FDLR-Net [100] Feature misalignment Localization refinement module (LRM) SME-Net [101] Feature misalignment Feature splitting and merging module (FSM) GLNet [102] Feature misalignment MSP+SA HyNet [103] Feature misalignment Learning hyperscale feature representations ASSD [104] Feature misalignment Pseudo-Anchor Proposal Module (PAPM) VistrongerDet [105] Feature misalignment FPN-level, ROI-level, and head-level enhancements HRDNet [106] Feature misalignment MD-IPN+MS-FPN GLGCNet [107] Restricted local information Saliency enhancement modules DCL-Net [108] Restricted local information RFAM+PAM SMSR [109] Restricted local information Aggregating features learned at different depths GLSAN [24] Crowded targets GLDN+SARSA+LSRN GDF-Net [110] Crowded targets Global density model (GDM) GSNet [111] Crowded targets Feat...…”

Enhancing Real-time Target Detection in Smart Cities: YOLOv8-DSAF Insights

“…The implementation of FEM involves introducing additional attention mechanisms and feature enhancement at speci c layers of the network. Speci cally, FEM enhances the representation capability of the network at certain levels, directing more attention to the crucial features of the targets (Zhang & Shen, 2022). This helps improve the model's ability to identify targets in complex urban environments, such as distinguishing different types of vehicles and capturing pedestrian movements.…”

High-Resolution Network with Transformer Embedding Parallel Detection for Small Object Detection in Optical Remote Sensing Images

“…Han et al [40] designed a multi-scale receptive field enhancement module (MRFEM) for small objects. Zhang et al [41] proposed a multi-stage feature enhancement pyramid network to fuse features at varying scales for small objects with blurry edges. Popular attention mechanisms have also shown significant improvements in remote sensing detection.…”