基于轨迹重构与贝叶斯推理的空中机器人灯塔距离测绘技术

Distance measurement technology for aerial robot lighthouses based on trajectory reconstruction and Bayesian inference

摘要:

灯塔在目前仍然发挥着重大作用,其在船舶出港、返港和近岸航行时是出海人员非常依赖的建筑物。传统方法无人机导航受到多路径以及动态环境的干扰,使得其导航出现误差。为精准确认灯塔的工作范围,提出一种基于轨迹重构与贝叶斯推理的空中机器人灯塔距离测绘技术。设计基于捷联式惯性导航系统的惯性轨迹重构技术,实现空中机器人灯塔距离测绘中的惯性导航,实现更加精准的灯塔距离测绘。采用的空中机器人为千巡翼Q10复合翼无人机飞行平台,对其实施惯性导航,采集灯塔灯光距离测绘数据。设计基于贝叶斯推理的多线索视觉注意模型,获取灯塔距离测绘显著图。根据该视觉注意显著图搭建灯塔距离测绘中的航拍摄像机成像模型,计算灯塔灯光照射距离。测试结果表明,该技术在12次测试中,无人机平均导航误差最终仅为0.132mm,平均灯塔灯光距离测绘时间较短,多次测试后平均灯塔灯光距离测绘误差仅为27.41cm,整体测绘误差小。

Abstract:

Lighthouses still play a significant role at present, as they are highly relied on by seafarers when ships depart, return, and navigate nearshore. The traditional method of unmanned aerial vehicle navigation is affected by multipath and dynamic environment interference, resulting in navigation errors. To accurately confirm the working range of lighthouses, a distance measurement technology for aerial robot lighthouses based on trajectory reconstruction and Bayesian inference is proposed. Design inertial trajectory reconstruction technology based on strapdown inertial navigation system to achieve inertial navigation in aerial robot lighthouse distance measurement, and achieve more accurate lighthouse distance measurement. The aerial robot used is the Qianxunyi Q10 composite wing unmanned aerial vehicle flight platform, which performs inertial navigation and collects lighthouse light distance measurement data. Design a multi clue visual attention model based on Bayesian inference to obtain significant maps of lighthouse distance mapping. Build an aerial camera imaging model for lighthouse distance mapping based on the visual attention saliency map, and calculate the lighthouse lighting illumination distance. The test results show that in 12 tests, the average navigation error of the drone is only 0.132mm, and the average lighthouse light distance measurement time is relatively short. After multiple tests, the average lighthouse light distance measurement error is only 27.41cm, indicating a small overall measurement error.