基于Hopfield网络的建筑物航测路径自动规划方法

Automatic path planning method of UAV based on Hopfield network for building mapping mission

针对当前无人机航线规划算法对地物类型(植被、建筑物等)不敏感,对建筑物测图任务不具备针对性等问题,提出一种基于Hopfield网络的建筑物航测路径自动规划方法。首先利用归一化植被系数(NDVI)排除测区内的植被;然后结合无人机位姿和多视图立体几何,粗略估计测区内的地势高低,确定建筑物区域的位置;再利用改良的8方向检测算子估计建筑物的主方向,优化航线方向;最后构建覆盖多个建筑物区域的Hopfield网络计算全局的最短路径。实验表明:对比常规“条带状”航线,所提方法对建筑物航测的作业距离降幅为68.4%,作业效率提升为50.7%,证明了所提方法具有更高的效率。

An automatic path planning method of unmanned aerial vehicle(UAV)based on the Hopfield network was proposed to solve the problems of the poor ability and specificity of common algorithms in building mapping missions.Firstly,the normalized differential vegetation index(NDVI)was applied to locate the vegetation.Then the terrain elevation was estimated roughly by the multiple view stereo approach combining the position of the UAV,and the building regions were extracted by integrating the NDVI results.And then an improved 8-orientated operator was utilized to estimate the dominant direction,which optimized the route direction.Finally,a Hopfield network was constructed to iteratively solve the shortest path covering all the building regions.The experimental results showed that the flight distance of the proposed algorithm was 68.4%lower than the strip routes,and the efficiency was 50.7%higher than the strip routes,indicating that the proposed algorithm had higher efficiency.