Abstract
Water target recognition is a critical challenge for the perception technology of unmanned surface vessels (USVs). In the application of USV, detection accuracy and the inference time both matter, while it is tough to strike a balance and single-frame water target detection behaves unstable in the video detection. To solve these problems, many strategies are applied to increase YOLOv4’s performance, including network pruning, the focal loss function, blank label training, and preprocessing with histogram normalization. The optimized detection method achieves a mean average precision (mAP) of 81.74% and a prediction speed of 26.77 frames per second (FPS), which meets the USV navigation requirements. To build the integrated USV-based system for water target recognition, a water target dataset containing 9936 images is created from offshore USV experiments in which the human-in-the-loop annotation and mosaic data augmentation methods are used. The issues of miss detection and false alarm can be considerably mitigated by cascading the Siamese-RPN tracking network, and the major color of a water target can be retrieved using a local contrast saliency color detection scheme. The system being tested is called “ME120” includes an embedded edge computing platform (Nvidia Jetson AGX Xavier). Finally, online dataset learning demonstrates the improved YOLOv4 achieves an increase of 66.98% in FPS at the cost of a decrease of 0.79% in mAP when compared with the original YOLOv4 and offline navigation experiments validate that our system achieves high recognition capability while maintaining a high degree of robustness.
Highlights
WORK The purpose of this work is to investigate the lightweight water target detection system based on unmanned surface vessels (USVs)
To begin, utilizing the human-in-the-loop technique, water target datasets comprising of 9936 images were established
Various techniques for optimizing the YOLOv4 algorithm were considered: 1) We pruned the backbone to increase inference speed to 26 frames per second (FPS) from 15 FPS; 2) We used the Focal Loss function to correct for sample imbalance, which resulted in a 0.33% increase in mean average precision (mAP); 3) We used the blank labels training method to reduce false alarms in real-world
Summary
Vs will be critical in the future water surface environment. It is useful in a variety of industries, including marine emergency rescue, logistics, water quality monitoring, hydrological surveying, marine environment mapping, and water ecological protection. This USV is capable of intelligent obstacle avoidance via real-time transmission of high-definition water surface footage. Several methods, including network pruning, focal loss function training, blank label training, and histogram normalization preprocessing, are used to improve the performance of YOLOv4; 2) A system was established by combining dataset configuration, training tricks, and a threestage scheme for water target detection, tracking, and color detection. This system has been validated in real-world environments.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
