Classification of airborne LiDAR point cloud data based on dual channel residual network
To improve the insufficient information circulation in the classification of airborne LiDAR point cloud data by traditional residual network,an airborne LiDAR point cloud data classification model,namely DP-Res-Net,based on dual-channel residual network is proposed.DP-ResNet adopts an encoding-decoding structure.In the encoding stage,two different forms of dual channel residual structure and parameter-free aggregation opera-tor are mainly combined.This can not only strengthen the circulation of network information,but also reduce network parameters.The decoding stage is done using traditional inverse distance weighting and 1×1 convo-lution.To verify the classification performance of the DP-ResNet model,classification experiments were per-formed on the GML DataSetA dataset.The results show that compared with the benchmark network Closerlook,the OA and AvgF1 of DP-ResNet model are improved by 6.25%and 15.45%respectively,indicating better classification performance.Compared with other models,DP-ResNet also has strong competitiveness.
airborne laser radarpoint cloud dataresidual networkDP-ResNet model
国家科技期刊平台
NSTL
万方数据
