Enhanced YOLO v3 Tiny Network for Real-Time Ship Detection From Visual Image

Hao Li,Jianbo Liu,Zhaoquan Gu,Lianbing Deng,Cheng Yang

doi:10.1109/access.2021.3053956

Abstract

Different from ship detection from synthetic aperture radar (SDSAR) and ship detection from spaceborne optical images (SDSOI), ship detection from visual image (SDVI) has better detection accuracy and real-time performance, which can be widely used in port management, cross-border ship detection, autonomous ship, safe navigation, and other real-time applications. In this paper, we proposed a new SDVI algorithm, named enhanced YOLO v3 tiny network for real-time ship detection. The algorithm can be used in video surveillance to realize the accurate classification and positioning of six types of ships (including ore carrier, bulk cargo carrier, general cargo ship, container ship, fishing boat, and passenger ship) in real-time. Based on the original YOLO v3 tiny network, we have made the following fine tunings. 1) The preset anchors trained on Seaship annotation data have the similar “dumpy” shape as the normal ships, helping the network to achieve faster and better training; 2) Convolution layer instead of max-pooling layer and expanding the channels of prediction network improve the small target detection ability of the algorithm. 3) Due to the problem that large-scale ships are easily disturbed by the onshore building, complex waves and light on the water surface, we introduced attention module named CBAM into the backbone network, which make the model more focused on the target. The detection accuracy of the proposed algorism is obviously better than that of the original YOLO v3 tiny work. Although it is slightly inferior to the Yolo v3 network, it has faster speed than Yolo v3. However, the proposed algorithm is a better trade-off between real-time performance and detection accuracy, and is more suitable for actual scenes. Compared with the SOAT algorithm in Z. Shao et al. (2020), our algorithm has a 9.6% improvement in mAP and a faster speed.

Highlights

Ship detection is a fundamental issue in the realization of automated fishery management, vessel traffic service, port management, and naval warfare
3) Due to the problem that large-scale ships are disturbed by the onshore building, complex waves and light on the water surface, we introduced attention module named convolutional block attention module (CBAM) into the backbone network, which make the model more focused on the target
Because of synthetic aperture radar (SAR) with broad field of vision and all-day and all-weather capability [1], ship detection from synthetic aperture radar (SDSAR) has attracted a lot of research

Summary

INTRODUCTION

Ship detection is a fundamental issue in the realization of automated fishery management, vessel traffic service, port management, and naval warfare. (1) Based on the Seaship dataset, we redefined a group of anchors Using these anchors, the performance of YOLO v3 tiny network in ship detection has been improved. (2) We optimized the backbone and prediction network of YOLO v3 tiny network After these optimizations, the detection performance for small-scale targets, such as fishing ship and passenger ship, has been significantly improved. Compared with SENet including only one max-pooling operation, CBAM uses both maxpooling and avg-pooling in two independent dimensions, so it can extract more abundant high-level features In this manuscript, our purpose is to train a ship detection network which has higher accuracy and better real-time than the SOAT method in [26], so as to solve the problem of effective and real-time supervision of illegal cross-border ships in Hengqin Island. Under the promise of ensuring the detection quality, the shorter the time, the better

EXPERIMENT ANALYSIS

Findings

CONCLUSION