四旋翼视觉区域定位系统设计  

A design of visual area localization system for quadrotor

作　　者：张鹏 沈捷[1] 王莉[1] 任重庚 

机构地区：[1]南京工业大学电气工程与控制科学学院,江苏南京211816

出　　处：《电子技术应用》2017年第2期44-46,50,共4页Application of Electronic Technique

摘　　要：提出四旋翼无人机利用机载摄像头实现视觉定位的方案。采用金字塔Lucas-Kanade稀疏光流方法来获取图像特征点的水平偏差。将高度、惯性和水平偏差进行数据融合,融合的数据作为位置式PID控制器的反馈输入。为了缩减视觉定位过程中轨迹检测的误差,使用前向后向误差去除误匹配点;对某时刻跟踪点丢失问题,使用离散卡尔曼滤波对该时刻的状态进行估计。仿真和实验表明,该方案能使得四旋翼无人机在一定高度下具有良好的定位效果。This paper proposes a visual localization strategy based on quadrotor with micro camera. The pyramid Lucas-Kanade sparse optical flow is applied to obtain the horizontal deviation of feature points. The data fusion of height, inertia, and horizontal deviation is used as the input of positional PID controller. To reduce the error of tracking trajectory, the forward-backward error algorithm is employed to eliminate the mismatching points. Kalman filtering algorithm is adopted to estimate the current state when feature points are failed to be tracked. The simulation and experiment show the effectiveness of proposed method within a certain range.

关 键 词：光流 数据融合 前向后向误差 PID 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]