基于改进型粒子群算法的曲面匹配与位姿获取  

Surface fitting and position-pose measurements based on an improved SA-PSO algorithm

作　　者：罗磊[1,2] 陈恳[1] 杜峰坡 马振书 

机构地区：[1]清华大学机械工程系,北京100084 [2]总装军械技术研究所,石家庄050000

出　　处：《清华大学学报（自然科学版）》2015年第10期1061-1066,共6页Journal of Tsinghua University(Science and Technology)

基　　金：国家"八六三"高技术项目(2009AA043701)

摘　　要：有限散乱点模式匹配是机器视觉与模式识别领域中的一个基础问题和重要环节,在目标识别、医学图像配准、遥感图像匹配、目标位姿获取等方面得到了广泛应用。该文提出了一种变概率密度分布的改进型模拟退火粒子群算法,能够加速算法收敛并搜索到全局最优解,提高匹配准确性。通过采集已知旋转体目标局部表面有限散乱点,构造适应度函数并进行坐标变换,实现散乱点与曲面的匹配和目标位姿获取,讨论了误差影响因素和算法适用性。应用结果表明:该方法与最小二乘曲面拟合方法相比,所需散乱点数目少,计算效率和精度高,且具有对散乱点采集误差不敏感等优点,能够满足排爆机器人抓取目标局部裸露情况下的位姿获取应用需求。Scattered point pattern matching is an important issue in computer vision and pattern recognition, which is widely used for target recognition, medical and remote sensing image registration, and position and pose measurements. This paper describes an algorithm for use in an explosive ordnance disposal robot for when the target object is partial exposed. A simulated annealing-particle swarm optimization （SA-PSO） algorithm based on particle density distribution changes for scattered point pattern matching is developed. The algorithm is more accurate and faster than previous algorithms. The fitness function is constructed from a series of scattered points collected from the local surface of a rotated object. Accurate surface fitting and pose parameters are found using coordinate transforms. The factors that affect the accuracy and applicability are discussed in detail. Tests show that this method is accurate and efficient, with less points needed and less sensitivity to point errors compared with the traditional least squares method.

关 键 词：粒子群算法 模拟退火 曲面匹配 排爆机器人 

分 类 号：TP242.3[自动化与计算机技术—检测技术与自动化装置]