基于深度学习的刚体点云配准变换参数预测  

Transformation Parameters Prediction of Rigid Body Point Cloud Registration Based on Deep Learning

作　　者：左文艳[1] 王明 

机构地区：[1]江苏联合职业技术学院镇江分院机电工程系,江苏镇江212016 [2]镇江市高等专科学校现代教育技术中心,江苏镇江212028

出　　处：《工业控制计算机》2022年第3期100-101,共2页Industrial Control Computer

基　　金：镇江市社会发展指导性项目(FZ2019061);镇江市高等专科学校校级科研项目(GZQNZX2019019)。

摘　　要：随着激光雷达、立体相机和结构光传感器等点云数据采集设备的发展,点云配准在计算机视觉和机器人技术领域引起了广泛关注。参数变换预测作为点云配准的关键步骤之一,寻找一个较好的预测方法可以有效地提高配准的精度。使用深度学习TensorFlow框架构建前馈神经网络,充分分析点云提取地全局特征,从全局特征中自适应学习变换参数,代替传统数学方法,为点云配准变换参数预测提供了一种参考手段,具有一定的实际意义。With the development of point cloud data acquisition equipment such as lidar,stereo cameras,and structured light sensors,point cloud registration has attracted widespread attention in the fields of computer vision and robotics.Transformation parameters prediction is one of the key steps of point cloud registration.Finding a better prediction method can effectively improve the accuracy of registration.This paper uses the deep learning TensorFlow framework to build a feedforward neural network,which fully analyzes the global features extracted from the point cloud,and adaptively learns the transformation parameters from the global features instead of traditional mathematical methods,and provides a reference method for the prediction of the transformation parameters of the point cloud registration,which has a certain practical significance.

关 键 词：深度学习 TensorFlow 前馈神经网络 点云配准 变换参数预测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP18[自动化与计算机技术—计算机科学与技术]