无人直升机自主着舰的计算机视觉技术  

Computer vision scheme for autonomous landing of unmanned helicopter on ship deck

作　　者：王晓剑[1] 潘顺良[1] 宋子善[1] 沈为群[1] 

机构地区：[1]北京航空航天大学自动化科学与电气工程学院,北京100083

出　　处：《北京航空航天大学学报》2007年第6期686-689,共4页Journal of Beijing University of Aeronautics and Astronautics

摘　　要：采用自主设计特征图案的方法,研究了一种基于视觉导引实现无人直升机自主着舰的快速算法.分析了制约位姿参数估计实时性的瓶颈,提出了一种基于信息分块的图像处理方法.引入选择性坐标变换的方法,获取图像处理的理想区域;采用直线方向粗识别的途径,确定参数区间,缩小了直线hough变换遍历空间,减小了底层图像处理的运算量.对真实图像的测试结果表明:对于768像素×576像素大小的图像帧,完成目标检测和位姿参数识别任务耗时小于30 ms,实现了视频流的实时处理;等效直升机与舰船相距10 m时,位置参数的均方根(RMS,Root Mean Square)误差在2cm内,姿态参数的RMS误差小于1.5°.算法能够满足自主着舰控制的实时性和实用性要求.A computer vision solution based on landing mark for guiding an unmanned helicopter to the ship deck was studied. Factors which limited the pose estimation efficiency were analyzed and a novel information-blocking based image processing method was proposed. With the optional coordinate transformation the ideal region for image processing was obtained, and by applying the technique of line direction coarse identification, the line parameters scope was determined and the space need to be traversed in line hough transformation was narrowed. The computation of low layer image processing was reduced. Tests on real image show that the target detection and state estimations in a 768 576 pixel size image may be completed within 30 ms. The state estimations are accurate to within 2 cm in each axis of translation and 1.5 o in each axis of rotation when the equivalent distance between the helicopter and the ship deck is about 10 m, and the algorithm is effective for the control of unmanned helicopter autonomous landing on ship deck.

关 键 词：无人飞行器 自主导航 计算机视觉 HOUGH变换 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]