基于视觉的AUV末端回收导引方法研究

Research on AUV terminal recovery Guiding method based on vision

摘要:

水下自主无人潜航器(AUV,Autonomous Underwater Vehicle)由于其具备高隐蔽性,超机动性等特点,是如今探索海洋的重要工具。作为自主机器人,AUV如何自主定位回收至坞站是研究中必不可少的环节。依托于实验室独创的四桨无舵矢量推进型AUV,通过改进Canny边缘检测,采用自适应阈值方法动态调整轮廓,在阈值最优时采用最小外接圆方法确定圆心。经Unity3D仿真和水池试验可知,该方法简单实用,鲁棒性强,且相比于传统的AUV形式以及传统的图像识别方法,该新型AUV在基于单目视觉的自适应阈值分割检测方法下,水下末端导引对接精度(优于20 cm)和对接成功率(大于80%)上均得到了大幅度的提高,在实际应用中对 AUV能源补给、数据下载 /上传、设备检修等具有重要的应用价值。

Abstract:

The Autonomous Underwater Vehicle (AUV) is a crucial tool for ocean exploration due to its exceptional concealment capabilities and supermobility. Locating and retrieving the AUV to the docking station is an essential aspect of research in autonomous robotics. Building upon the laboratory"s original four-paddle ruddless vector-propelled AUV, we have enhanced the contour by refining Canny edge detection techniques. Additionally, we determine the circle center using the minimum circumferential circle method at optimal thresholds. Through Unity3D simulation and pool testing, it has been demonstrated that this approach is straightforward, practical, and robust. In comparison to traditional AUV designs and conventional image recognition methods, this novel AUV can effectively employ an adaptive threshold segmentation detection method based on monocular vision. Consequently, significant improvements have been achieved in terms of underwater terminal guidance docking accuracy (better than 20 cm) as well as docking success rate (exceeding 80%). These advancements hold substantial application value for AUV energy supply management, data download/upload operations, and equipment maintenance in real-world scenarios.