【摘要】 针对电力杆塔无人机智能巡检手动选取特征点对专业要求高、效率低、点位命名不统一等问题，该文提出一种基于高密度点云与杆塔模型属性信息的特征点智能提取方法。首先将杆塔点云数据与杆塔模型属性信息通过模型名称建立联系，得到杆塔关键点，然后利用特征点预测技术得到特征点粗略位置，最后通过特征点智能提取技术得到特征点准确位置。基于4类杆塔模型的实验结果表明：该方法的特征点提取平均正确率在88.1%以上，大幅减少了人工干预，提高了提取效率，按顺序编码命名的特征点为任务规划带来了便捷。通过实践应用，平均效率比手动规划提高3.3倍，巡检照片满足技术要求，在实际电网巡检工程中得到很好的应用。更多还原

【Abstract】 Aiming at the problems of high professional requirements, low efficiency and inconsistent feature points naming of manual selection of feature points in unmanned aerial vehicle(UAV) intelligent inspection of power tower. An intelligent extraction method of feature points based on high density point cloud data and tower model attribute information was proposed. Firstly, the tower point cloud data was connected with the tower model attribute information through the model name to obtain the tower key points, and then the rough position of the feature points was obtained by using the feature point prediction technology. Finally, the accurate position of feature points was obtained by feature point intelligent extraction technology. The experimental results based on four type tower models showed that the average accuracy of feature point extraction of this method was more than 88.1%,which greatly reduced manual intervention and improved the extraction efficiency. The feature points encoded and named in sequence brought convenience to task planning. Through practical application, the average efficiency was 3.3 times higher than that of manual planning in task planning, and the patrol photos met the technical requirements. This method has been well applied in the actual power tower inspection project.更多还原