The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. ex. Some numerals are expressed as "XNUMX".
Copyrights notice
The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. Copyrights notice
Como uma espécie de veículo marítimo, os Veículos de Superfície Não Tripulados (USV) são amplamente utilizados nos campos militar e civil devido ao seu baixo custo, boa ocultação, forte mobilidade e alta velocidade. A detecção de obstáculos de alta precisão desempenha um papel importante na navegação autônoma do USV, o que garante seu posterior planejamento de trajetória. A fim de melhorar ainda mais o desempenho da detecção de obstáculos, propomos uma arquitetura codificador-decodificador chamada Fusion Refinement Network (FRN). A parte do codificador com uma estrutura de rede mais profunda permite extrair recursos visuais mais ricos. Em particular, uma camada de convolução dilatada é usada no codificador para obter uma grande variedade de características de obstáculos em ambientes marinhos complexos. A parte do decodificador atinge a fusão de recursos de múltiplos caminhos. Módulos de refinamento de atenção (ARM) são adicionados para otimizar recursos, e um algoritmo de fusão que pode ser aprendido chamado Módulo de fusão de recursos (FFM) é usado para fundir informações visuais. Os resultados da validação experimental em três conjuntos de dados diferentes com imagens marinhas reais mostram que o FRN é superior às redes de segmentação semântica de última geração na avaliação de desempenho. E o MIoU e o MPA do FRN podem atingir um pico de 97.01% e 98.37%, respectivamente. Além disso, o FRN conseguiu manter uma alta precisão com apenas 27.67 milhões de parâmetros, o que é muito menor do que a mais recente rede de detecção de obstáculos (WaSR) para USV.
Weina ZHOU
Shanghai Maritime University
Xinxin HUANG
Shanghai Maritime University
Xiaoyang ZENG
Fudan University
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copiar
Weina ZHOU, Xinxin HUANG, Xiaoyang ZENG, "Obstacle Detection for Unmanned Surface Vehicles by Fusion Refinement Network" in IEICE TRANSACTIONS on Information,
vol. E105-D, no. 8, pp. 1393-1400, August 2022, doi: 10.1587/transinf.2021EDP7254.
Abstract: As a kind of marine vehicles, Unmanned Surface Vehicles (USV) are widely used in military and civilian fields because of their low cost, good concealment, strong mobility and high speed. High-precision detection of obstacles plays an important role in USV autonomous navigation, which ensures its subsequent path planning. In order to further improve obstacle detection performance, we propose an encoder-decoder architecture named Fusion Refinement Network (FRN). The encoder part with a deeper network structure enables it to extract more rich visual features. In particular, a dilated convolution layer is used in the encoder for obtaining a large range of obstacle features in complex marine environment. The decoder part achieves the multiple path feature fusion. Attention Refinement Modules (ARM) are added to optimize features, and a learnable fusion algorithm called Feature Fusion Module (FFM) is used to fuse visual information. Experimental validation results on three different datasets with real marine images show that FRN is superior to state-of-the-art semantic segmentation networks in performance evaluation. And the MIoU and MPA of the FRN can peak at 97.01% and 98.37% respectively. Moreover, FRN could maintain a high accuracy with only 27.67M parameters, which is much smaller than the latest obstacle detection network (WaSR) for USV.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.2021EDP7254/_p
Copiar
@ARTICLE{e105-d_8_1393,
author={Weina ZHOU, Xinxin HUANG, Xiaoyang ZENG, },
journal={IEICE TRANSACTIONS on Information},
title={Obstacle Detection for Unmanned Surface Vehicles by Fusion Refinement Network},
year={2022},
volume={E105-D},
number={8},
pages={1393-1400},
abstract={As a kind of marine vehicles, Unmanned Surface Vehicles (USV) are widely used in military and civilian fields because of their low cost, good concealment, strong mobility and high speed. High-precision detection of obstacles plays an important role in USV autonomous navigation, which ensures its subsequent path planning. In order to further improve obstacle detection performance, we propose an encoder-decoder architecture named Fusion Refinement Network (FRN). The encoder part with a deeper network structure enables it to extract more rich visual features. In particular, a dilated convolution layer is used in the encoder for obtaining a large range of obstacle features in complex marine environment. The decoder part achieves the multiple path feature fusion. Attention Refinement Modules (ARM) are added to optimize features, and a learnable fusion algorithm called Feature Fusion Module (FFM) is used to fuse visual information. Experimental validation results on three different datasets with real marine images show that FRN is superior to state-of-the-art semantic segmentation networks in performance evaluation. And the MIoU and MPA of the FRN can peak at 97.01% and 98.37% respectively. Moreover, FRN could maintain a high accuracy with only 27.67M parameters, which is much smaller than the latest obstacle detection network (WaSR) for USV.},
keywords={},
doi={10.1587/transinf.2021EDP7254},
ISSN={1745-1361},
month={August},}
Copiar
TY - JOUR
TI - Obstacle Detection for Unmanned Surface Vehicles by Fusion Refinement Network
T2 - IEICE TRANSACTIONS on Information
SP - 1393
EP - 1400
AU - Weina ZHOU
AU - Xinxin HUANG
AU - Xiaoyang ZENG
PY - 2022
DO - 10.1587/transinf.2021EDP7254
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E105-D
IS - 8
JA - IEICE TRANSACTIONS on Information
Y1 - August 2022
AB - As a kind of marine vehicles, Unmanned Surface Vehicles (USV) are widely used in military and civilian fields because of their low cost, good concealment, strong mobility and high speed. High-precision detection of obstacles plays an important role in USV autonomous navigation, which ensures its subsequent path planning. In order to further improve obstacle detection performance, we propose an encoder-decoder architecture named Fusion Refinement Network (FRN). The encoder part with a deeper network structure enables it to extract more rich visual features. In particular, a dilated convolution layer is used in the encoder for obtaining a large range of obstacle features in complex marine environment. The decoder part achieves the multiple path feature fusion. Attention Refinement Modules (ARM) are added to optimize features, and a learnable fusion algorithm called Feature Fusion Module (FFM) is used to fuse visual information. Experimental validation results on three different datasets with real marine images show that FRN is superior to state-of-the-art semantic segmentation networks in performance evaluation. And the MIoU and MPA of the FRN can peak at 97.01% and 98.37% respectively. Moreover, FRN could maintain a high accuracy with only 27.67M parameters, which is much smaller than the latest obstacle detection network (WaSR) for USV.
ER -