Related Experiment Video
Updated: May 28, 2025
08:25
Combining Eye-tracking Data with an Analysis of Video Content from Free-viewing a Video of a Walk in an Urban Park Environment
Published on: May 7, 2019
8.9K
Lightweight CNN-Based Visual Perception Method for Assessing Local Environment Complexity of Unmanned Surface Vehicle
Tulin Li1, Xiufeng Zhang1, Yingbo Huang1
1Yunnan Key Laboratory of Intelligent Control and Application, Faculty of Mechanical and Electrical Engineering, Kunming University of Science and Technology, Kunming 650500, China.
Sensors (Basel, Switzerland)
|February 13, 2025
Summary
This study introduces a visual perception method for unmanned surface vehicles (USVs) using a lightweight convolutional neural network (CNN) and heading angle. It enhances environmental detection for optimized path planning, improving efficiency by 21%.
Area of Science:
- Robotics
- Artificial Intelligence
- Computer Vision
Background:
- Unmanned Surface Vehicles (USVs) require effective environmental detection for optimized path planning.
- Existing heuristic algorithms struggle with inadequate environmental perception.
- Integrating real-time environmental data is crucial for enhancing USV navigation.
Purpose of the Study:
- To develop a comprehensive visual perception method for USVs.
- To improve the environmental detection capabilities of heuristic algorithms.
- To enhance the path planning efficiency and performance of USVs.
Main Methods:
- A lightweight convolutional neural network (CNN) with residual learning blocks was employed.
- The CNN utilized multi-feature inputs, including local environmental images and real-time heading angle.
- Human expertise was incorporated via a majority voting system for intuitive label classification.
Main Results:
- The developed model achieved an 80% reduction in size while maintaining over 90% accuracy.
- Environmental recognition capability of the heuristic algorithm was significantly improved.
- Overall USV path planning performance increased by approximately 21%.
Conclusions:
- The proposed method offers a human-like comprehensive perception ability for USVs.
- This approach enhances the heuristic algorithm's environmental recognition and path planning efficiency.
- The method provides a robust solution for optimizing USV search and navigation.
