基于格子波尔兹曼的机器人视觉特征点检测

针对Harris特征点检测算法中图像存在角点信息丢失的问题，提出基于格子波尔兹曼结合Harris算法的机器人视觉特征点检测方法。格子波尔兹曼具有并行操作的优点，可以保证机器人导航的实时性。通过格子波尔兹曼方法对图像进行预处理，然后使用Harris算法，实现特征点的有效检测。实验结果表明，该方法更加准确的选择特征点，减少特征点数目的同时，提升了特征点的质量，完成了特征相对不明显的点的检测。

Detection and extraction of feature points in robot vision using Lattice Boltzmman method

This paper proposes a new method of feature point detection and extraction in robot vision. The method is based on the combined use of Lattice Boltzmman method (LBM) and Harris algorithm, allowing coping with the problem of corner point missing in feature detection and extraction. The LBM has the advantage of being able to be handled using parallel computation, thus ensuring the real-time navigation possibility of robots. The results showed that the LBM-based preprocessing, followed by Harris algorithm, enables us to more accurately select feature points, reduce the number of necessary feature points, improve the quality of selected feature points, and extract difficult feature points.