基于数字高程地图的无人机备降场自主搜寻

Potential Landing Sites Research of UAV through Digital Elevation Model

随着无人机在现代社会各领域的不断拓展,用户对于无人机自主决策能力的智能化需求日益增长。与此同时,无人机的应用场景也逐渐从传统的室内与城市环境,扩展至地图尺度更大,环境更为陌生的野外场景。本研究针对野外场景下,无人机备降场分割面临环境障碍影响较大的问题,利用数字高程地图作为地图数据来源,设计了基于坡度的传统备降场分割算法,以及基于神经网络的备降场分割算法。在实验阶段,本研究对两种算法的计算速度与准确率进行比较。实验结果表明,本研究设计的基于神经网络的备降场分割算法在稳定性和计算速度方面具有显著优势,适用于硬件性能受限的无人机。

The application of Unmanned Aerial Vehicles(UAVs)has seen a significant increase in recent years,with a growing demand for automatic UAV manipulations in various applications.As a result,the application of UAVs has expanded beyond traditional indoor and urban scenes to larger-scale field environments,presenting new challenges in terms of potential landing site segmentation.To address this issue,this research proposes a traditional gradient-based potential landing site segmentation algorithm is proposed along with a neural network-based approach that leverages Digital Elevation Models(DEMs)as a map source.The study compares the calculation speed and accuracy of the two algorithms in experiments.Results show that the neural network-based algorithm significantly improves stability and calculation speed compared to traditional segmentation approaches,making it a suitable solution for limited hardware performance requirements of UAVs.