无人机自动着陆中的机器视觉辅助技术

设计了一套辅助无人机自动着陆的机器视觉系统。该系统由机载硬件设备和用户开发软件共同组成,用以完成数字图像处理任务和无人机运动参数估计任务。系统传感器包括一个单目摄像机和机载惯性陀螺;数字图像处理使用的主要算法有图像的轮廓提取、角点检测和模版匹配。基于角点处各个方向上灰度差变化较大的特征,依据最小核值相似(SUSAN)算法和角点几何结构分析,提出一种改进的角点特征提取算法;根据任务开发的位置参数估计算法依据摄像机透视投影理论,运用摄像机成像标定方法导出了一种高精度的位置测量模型。通过计算机仿真表明,所提出的计算机视觉位置参数估计算法可以达到无人机着陆过程的精度要求。

Computer Vision Assisted Autonomous Landing of UAV

An analysis of the orientation error for a computer vision assisted integrated navigation scheme for an autonomous landing maneuver of unmanned aerial vehicle(UAV) is presented.The system utilizes the binocular on-board camera fixed on the optoelectronic tracking platform as the sensor to obtain the plan position information.The image processing of the landing system uses such algorithms as template-matching,contour-extraction and feature-point-detection.Since the corners are image points showing strong two dimensional intensity changes while well dis-tinguished from nearby points,the paper establishes a new fast and efficient corner detector based on SUSAN and the feature of corner structure to achieve the corner point detecting task.An algorithm is offered to obtain flight position information based on the knowledge of analytic geometry.The measurement model for the automatic landing system of UAV is deduced according to camera perspective projection theory.Computer simulation results indicate that the navigation scheme could operate properly for the autonomous landing of the UAV.