城市场景重访车载点云位置一致性改正

Position Consistency Correction of Revisit Mobile Laser Scanning Point Cloud in Urban Scene

车载激光点云在城市道路资产管理、高精驾驶地图、农村宅基地调查、高速公路改扩建、智能交通等国家重大工程应用中发挥着非常重要的作用;然而,受道路环境复杂、定位信号受遮挡、定姿误差累积等影响,导致往返或不同时相的重访车载点云存在分米甚至米级的位置偏差,严重影响后续数据处理与应用。为解决上述技术瓶颈,本文提出一种城市场景重访车载点云位置一致性改正算法。首先,依据车载轨迹的加速度与角速度将车载点云数据进行层次化分段,同时保证重访段的重叠度;然后,提取分段内的二进制形状上下文(Binary Shape Context, BSC)特征,并依据视觉单词与先验信息加速同名BSC特征匹配;最后,依次进行重访粗分段和细分段点云的两两配准,并剔除不可靠的两两配准结果。实验表明,本文方法能有效改正城市场景重访车载点云中的位置不一致问题,对于不同偏差级别和时相的车载点云,具有很高的鲁棒性和时间效率。

The Mobile Laser Scanning(MLS)point clouds play a very important role in national major engi-neering applications such as urban road asset management,high-definition driving map,rural homestead survey,highway reconstruction and expansion.However,owing to complex road envi-ronment,occluded positioning signal and time-accumulation of attitude error,the MLS point clouds collected by the back and forth scans or among multiple excursions in the same region often suffer misalignment ranging from sub-meter to meters,which impedes the subsequent processing and ap-plications.To deal with the technical bottleneck mentioned above,a method of MLS point cloud po-sition consistency correction in urban scene is proposed.Firstly,the MLS point clouds are divided into multi-scale sub-regions based on the acceleration and angular velocity of each trajectory point,and the overlap degree of revisited sub-regions is ensured at the same time.Secondly,binary shape context(BSC)features in sub-regions are extracted,and visual words and prior information are used to accelerate feature matching.Thirdly,pairwise registration of large and small revisited sub-regions is carried out in turn,and unreliable registration results are removed.The perfor-mance of the proposed method is evaluated on several challenging MLS point clouds with different deviation levels and different temporal,showing good robustness,accuracies and efficiencies.