摘要
研究一种结合ORB特征点的改进型RatSLAM算法,通过设置局部自适应阈值对图像的像素分类,以此进行特征点的筛选与提取,简化特征点的冗杂与聚集。再使用渐进一致采样算法(PROSAC)进行特征点匹配。通过与传统Rat⁃SLAM算法和使用了SURF特征点的RatSLAM算法进行对比分析,仿真实验的结果验证了改进后算法具有快速性与低误匹配率。再使用Ubuntu中的ROS平台搭建仿真模拟环境,让机器人在环境中运动通过坐标化进行认知地图的构建。通过视觉模板匹配检测图像的相似度进行闭环检测和多圈运动的轨迹比较,仿真实验的结果验证了构建的认知地图具有更高的定位精度,且在长时间的运行任务中累计误差得到了更好的校正。
The thesis studies an improved RatSLAM algorithm that combines ORB feature points.By setting local adaptive thresholds,the pixels of the image are classified,so as to screen and extract feature points,simplifying the redundancy and aggregation of feature points.Then use the progressive consensus sampling algorithm(PROSAC)for feature point matching.Through comparative analysis with the traditional Rat⁃SLAM algorithm and the RatSLAM algorithm that uses SURF feature points,the results of simulation experiments verify that the improved algorithm has rapidity and low mismatch rate.Then use the ROS platform in Ubuntu to build a simulation environment,and let the robot move in the environment to construct a cognitive map through coordinated.Through the visual template matching to detect the similarity of the image,the closed-loop detection and the multi-circle motion trajectory are compared.The results of the simulation experiment verify that the constructed cognitive map has higher positioning accuracy,and the cumulative error in the long-term running task is obtained bet⁃ter correction.
作者
洪涛
史涛
任红格
HONG Tao;SHI Tao;REN Hongge(School of Electrical Engineering,North China University of Technology,Tangshan 063210;School of Electronic Engineering,Tianjin University of Technology,Tianjin 300384;School of Control and Mechanical Engineering,Tianjin Urban Construction University,Tianjin 300384)
出处
《现代计算机》
2021年第21期47-52,共6页
Modern Computer
基金
河北省自然科学基金项目(No.F2018209289)。
