基于局部熵的SLAM视觉里程计优化算法

Optimization Algorithm of Visual Odometry for SLAM Based on Local Image Entropy

针对移动机器人视觉同步定位与地图创建中由于相机大角度转动造成的帧间匹配失败以及跟踪丢失等问题,提出了一种基于局部图像熵的细节增强视觉里程计优化算法.建立图像金字塔,划分图像块进行均匀化特征提取,根据图像块的信息熵判断其信息量大小,将对比度低以及梯度变化小的图像块进行删除,减小图像特征点计算量.对保留的图像块进行亮度自适应调整,增强局部图像细节,尽可能多地提取能够表征图像信息的局部特征点作为相邻帧匹配以及关键帧匹配的关联依据.结合姿态图优化方法对位姿累计误差进行局部和全局优化,进一步提高移动机器人系统性能.采用TUM数据集测试验证,由于提取了更能反映物体纹理以及形状的特征属性,本文算法的运动跟踪成功率最高可提升至60%以上,并且测量的轨迹误差、平移误差以及转动误差都有所降低.与目前ORB-SLAM2系统相比,本文提出的算法不但提高了移动机器人视觉定位精度,而且满足实时SLAM的应用需要.

For the problems of failed matching and trackingloss due to big camera rotation in simultaneous localizationand mapping(SLAM)for mobile robots,an optimized detailenhancement algorithm of visual odometry based on the localimage entropy is proposed.The image pyramid is built andis divided into blocks on each level to extract features homo-geneously.The information of each image block is determinedthrough its entropy value and the blocks with low contrast andsmall intensity gradient will be deleted to reduce feature calcu-lation.Nonlinear and adaptive illumination adjustment on eachreserved block is applied to increase local image details.Lo-cal features that representing image information is preserved asmuch as possible to be the correlations between adjacent framesand keyframes.Combined with the pose graph optimizationmethod,the local and global optimization of accumulation error is carried out to further improve the system performancefor mobile robot.The proposed method is verified on the TUMdataset.Since using the feature properties which are more reflective of texture and shape,the maximum success rate of motiontracking is increased to over 60%.And the results also showthat the tracking error,translational error and rotation error isdecreased.Compared with the original system ORB-SLAM2,this method can not only improve visual positioning accuracy ofthe mobile robot,but also meet the application need of realtime SLAM requirement.