基于语义特征的3D点云室内目标检测

3D Point Cloud Indoor Object Detection Based on Semantic Features

当前有许多关于点云目标检测的研究,但这些研究大多适用于室外环境。室内目标相对于室外目标的检测难点在于体积较小且细节较多。而点云数据具有稀疏性,不利于检测小型目标。为了提高室内目标的检测精度,提出了一种改进VoteNet的目标检测模型。首先,该模型添加了一个能有效提取语义特征的前置网络,通过将提取到的语义特征与几何特征进行融合,得到更有效的特征块用于目标检测。其次,通过在VoteNet投票簇生成阶段添加语义约束,一定程度上解决了VoteNet投票簇中点的语义非一致性问题。在公开数据集ScanNet上进行实验,实验证明改进模型在检测室内目标时获得了更高的检测精度。

There are many researches on point cloud object detection,but most of these researches are applicable to outdoor environments.The difficulty of detecting indoor objects relative to outdoor objects is that they are smaller in size and more detailed.The point cloud data is sparse,which is not conducive to detecting small objects.In order to improve the detection accuracy of in⁃door objects,an improved VoteNet object detection model is proposed.Firstly,the model adds a pre-network that can effectively ex⁃tract semantic features.By fusing the extracted semantic features with geometric features,more effective feature blocks are obtained for target detection.Secondly,by adding semantic constraints in the generation stage of VoteNet voting clusters,the problem of se⁃mantic inconsistency in the midpoints of VoteNet voting clusters is solved to a certain extent.Experiments are carried out on the pub⁃lic data set ScanNet,and the experiment proves that the improved model obtained higher detection accuracy when detecting indoor objects.