基于DBSCAN与梯度划分的Kinect障碍物轮廓检测算法

针对移动机器人的环境检测和避障问题中传感器获取的信息不够全面及准确,无法准确提供周围环境信息等问题,文中提出了利用Kinect传感器来获取周围环境的色彩信息和深度数据,并且提出了一种利用梯度划分和DBSCAN聚类方法来处理Kinect传感器获得的深度数据图。该算法首先使用梯度障碍物边缘检测方法对Kinect获取得到的深度图进行快速高效的处理障碍物边缘轮廓,并对算法中的差分参数进行改进,使得计算得到的梯度结果更准确。然后对比不同的聚类方法,使用BDSCAN聚类方法来对检测划分完毕的障碍物进行聚类分析,最后通过安排具体实验对该算法进行验证。实验结果表明,该算法能够对周围环境障碍物进行准确划分,可行区域效果明显,对不同物体的成功检测率较高,验证了算法的有效性。

Kinect Obstacle Detection Algorithm Based on DBSCAN and Aradient Partition

For the environment detection and obstacle avoidance problems of mobile robots, the information acquired by the sensors was not comprehensive and accurate, and it was impossible to accurately provide information about the surrounding environment. In this paper, the use of Kinect sensors to obtain the color information and depth data of the surrounding environment was proposed. A depth data map obtained by the Kinect sensor using gradient partitioning and DBSCAN clustering method. Firstly, the gradient obstacle edge detection method was used to quickly and efficiently process the edge contour of the obstacle obtained by Kinect, and the difference parameters in the algorithm are improved, so that the calculated gradient result was more accurate. Then compare the different clustering methods, use BDSCAN clustering method to cluster the detected obstacles, and finally verify the algorithm by arranging specific experiments. The experimental results showed that the algorithm can accurately divide the surrounding obstacles, and the feasible area was effective. The successful detection rate for different objects was more accurate. The validity of the algorithm was verified.