基于体素特征融合的FOD目标智能认知  

Intelligent Cognition of FOD Based on Voxel Feature Fusion

作　　者：褚昕悦 赵旭[1] 李连鹏[1] 刘文 代牮 Chu Xinyue;Zhao Xu;Li Lianpeng;Liu Wen;Dai Jian(Beijing Key Laboratory of High Dynamic Navigation Technology,Beijing Information Science&Technological University,Beijing 100192,China)

机构地区：[1]北京信息科技大学高动态导航技术北京市重点实验室,北京100192

出　　处：《应用激光》2023年第8期151-158,共8页Applied Laser

基　　金：国家重点研发计划课题(2020YFC1511702);北京市自然科学基金(4214071);北京学者计划。

摘　　要：机场跑道异物(FOD)的快速检测对于飞机的安全行驶至关重要,针对传统FOD检测技术中存在小尺寸FOD目标识别困难和实时性差等问题,提出一种基于体素特征融合的FOD检测方法。该方法首先提取感兴趣区域内的点云,依据随机抽样一致性算法(RANSAC)对地面点云进行校准;采用体素特征融合算法,对校准后的点云进行体素划分,并将每个体素内所有点云的平均Z值与平均反射率按一定比例进行特征融合,形成新的点云,以此将稠密点云在突出FOD的情况下进行减量;接着对地面点云与地面上FOD点云进行分割;最后采用基于KD-tree的欧式聚类算法对分割后地面上FOD点云进行聚类。试验结果表明,此方法可实现30 m远长宽高大于1.3 cm的FOD检测,比传统方法在探测FOD大小的性能上提高了35%,为激光雷达在FOD检测领域的应用提供了参考。The rapid detection of foreign object debris(FOD)is very important for the safe driving of aircraft.Considering the problems of small size FOD target recognition difficulty and poor real-time performance in traditional FOD detection technology,a FOD detection method based on voxel feature fusion is proposed.The method firstly extracts the point cloud in the region of interest and calibrates the ground point cloud according to the RANSAC method.Secondly,the voxel feature fusion algorithm is used to divide the calibrated point cloud into voxels,and the average Z value and the average reflectivity of all point clouds are fused in each voxel according to a certain proportion to form a new point.In this way,the dense point cloud is reduced in the case of protruding FOD.The ground point cloud and the point cloud of FOD on the ground are segmented.Finally,the Euclidean clustering is used based on the KD-tree algorithm and clusters the point cloud of the FOD on the ground after segmentation.The experimental results show that this method can realize FOD detection with length,width,and height greater than 1.3 cm in a distance of 30 meters.Compared with the traditional method,the detection FOD size performance increased by 35%,which provides a reference for lidar in FOD detection.

关 键 词：机场跑道异物 体素特征融合 随机抽样一致性算法 欧式聚类 激光雷达 

分 类 号：TN958.98[电子电信—信号与信息处理] TP391.41[电子电信—信息与通信工程]