复杂环境下基于Kinect的视觉定位算法研究

Research on Vision Location Algorithm Based on Kinect in Complex Condition

针对复杂环境下机器人的定位问题,研究基于Kinect传感器的视觉定位算法。对于机器人在运动过程中采集的多帧连续图像,利用SURF算法进行特征提取与匹配,并对彩色图像和深度图像进行配准对齐,采用RANSAC算法完成初步配准,剔除一些误匹配点,实现帧间图像相对运动的粗略估计,在此基础上进一步采用ICP算法进行精确配准,得到优化的定位估计参数。实验结果表明,此方法能较好地实现复杂环境下机器人的视觉定位,并且可以得到场景的重建。

This paper,aiming at solving the location problem of robot in the complex condition,studies the vision location algorithm based on the Kinect sensor.Firstly,the SURF algorithm was used to extract and match features of the multi-frame continuous image captured by the robot in motion.Then,the color images and the depth images were aligned and registered preliminarily with RANSAC algorithm,eliminating some false matching points to achieve a rough estimate of the relative motion of frames.Based on which,the ICP algorithm was further taken to carry out accurate registration,hence the optimal location estimation parameters were finally obtained.The experimental results show that better vision location of robot in complex environment can be achieved and the scene can be reconstructed with this method.