The current markerless augmented reality(AR) algorithm has poor performance on 3D registration un-der various luminance and texture conditions. In order to enable the algorithm to overcome the drawbacks of robust problem in realistic environment, the LoG-PTAMM algorithm is proposed with Laplacian of Gaus-sian(LoG) operator. Firstly edges and details were intensified in AR scenario by filtering streaming image pyramids with LoG operator. Then the ZMSSD threshold of reprojection process had been set through the values of average of final accept score and average search length, since the frames contrast varied dramati-cally after LoG transformation. Besides a bilateral optical flow method was implemented for promoted matching precision. Finally anIls oriented relocalisation algorithm was implemented to search the Gaus-sian filtered video frames for camera pose recovery in case of tracking lost. Experimental performance was measured by average robustness index defined with hit rate and robustness, significant improvements was observed on illumination coherence and feature extraction in ordinary markerless AR scenarios in LoG-PTAMM algorithm comparing with PTAMM. Lifelong robust mapping and tracking in large scale en-vironment become possible under multiple luminance and texture conditions.%针对现有无标识增强现实算法容易受到光照条件以及环境纹理的影响，出现虚拟信息三维注册跟踪失败或者精度误差较大现象，致使鲁棒性无法得到保证的问题，提出一种基于 LoG 算子的无标识增强现实算法——LoG-PTAMM.首先使用 LoG 算子将输入视频帧所生成的图像金字塔滤波，突出场景中的边缘和细节；然后利用最终得分均值和平均搜索长度值确定重投影算法的ZMSSD阈值，以适应LoG变化后视频帧强烈的对比度变化，并使用双向光流法提升匹配精度；最后在跟踪丢失时采用以 Ils变量为核心的重定位算法查找经过高斯滤波处理的已有视频帧序列并恢复相机姿态.采用文中根据命中率与贡献度定义的鲁棒性指数均值度量的实验结果表明， LoG-PTAMM算法在室外多种无标识增强现实场景中具有光照一致性且易于提取大量鲁棒的环境特征点，可以在多种环境纹理或光照条件下进行长时间、大范围的建图和跟踪活动，鲁棒性较 PTAMM算法有较大程度地提升.
展开▼
