基于多传感器信息融合的机器人障碍物检测

Obstacle detection of robot based on multi-sensor information fusion

针对单一传感器在机器人避障过程中不能全面且准确定位障碍物的缺点,提出基于多传感器信息融合的障碍物检测方法。第一阶段使用视觉传感器检测未知环境中的障碍物,通过Zernike矩边缘检测方法提取障碍物图像边缘,然后采用Hough变换原理提取障碍物的直线特征,获得障碍物大概位置;第二阶段使用超声波传感器和红外传感器对障碍物进行检测,获得障碍物准确位置;最后使用联合卡尔曼滤波对3种传感器获得的信息进行融合,得出融合后的障碍物位置信息。实验结果表明:该方法克服视觉传感器、超声波传感器和红外传感器的局限性,可以精确感知机器人周围的未知环境信息,并能够检测和定位机器人路径上的障碍物,定位误差<6 cm,满足机器人避障的实时性和可靠性需求。

Aiming at the shortcoming that the single sensor could not locate the obstacle completely and accurately in the process of obstacle avoidance of the robot, an obstacle detection method based on multi-sensor information fusion was proposed. Firstly, vision sensor was used to detect the obstacle in unknown environment. The edge of the obstacle image was extracted by the Zernike moment edge detection method, then the Hough transform principle was used to extract the straight line feature of the obstacle, so as to obtain the approximate position of the obstacle. Secondly, ultrasonic sensor and infrared sensor were used to detect the obstacle to obtain the exact position of obstacles. Finally, the federated Kalman filter was used to fuse the information obtained by the three sensors to gain information of the obstacle position after fusion. The test result shows that this method can overcome the limitations of vision sensors, ultrasonic sensors and infrared sensors, and can accurately detect the unknown environmental information around the robot and detect and locate the obstacles on the path of robot with positioning error less than 6 cm, meeting the real-time and reliability of robot obstacle avoidance.