应用改进的粒子群优化模糊聚类实现点云数据的区域分割

发布时间：2018-06-24 09:33

本文选题：点云数据 + 区域分割　；参考：《光学精密工程》2017年04期

【摘要】：为实现点云数据的区域划分,提出一种基于改进的粒子群优化与模糊C-均值聚类的混合算法(SPSO-FCM算法)。针对在点云聚类过程中易过早捕获局部极小值的问题,算法首先用改进的粒子群算法——社会粒子群优化算法,对种群进行初始化,通过为每一个粒子设置不同的跟随阈值,来维护种群中个体多样性,加深对种群全局搜索的程度,避免陷入局部极小值;随后,设置种群中每个粒子当前最优位置和初始种群的最优位置,更新自由粒子的位置和跟随粒子的速度和位置;最后,采用模糊C-均值聚类算法求解隶属度矩阵,确定适应值函数,更新所有粒子的最优位置,并判断粒子和种群的位置优越性,得到准确的聚类中心,实现对点云数据的区域划分。以曲面复杂度不一致的点云模型为例对算法进行验证,探讨SPSO-FCM聚类算法的可行性,并与FCM聚类算法、遗传FCM聚类算法进行比对。实验结果显示,SPSOFCM聚类算法较其它两种算法,收敛速度快,迭代次数少,聚类准确,边界区域分割清晰,特别是对型面复杂、点云数据较多的机械零部件点云数据进行分割时,能得到更好的分割结果。
[Abstract]:A hybrid algorithm (SPSO-FCM) based on improved particle swarm optimization (PSO) and fuzzy C-means clustering (FCM) is proposed to realize the region partition of point cloud data. In order to solve the problem that local minimum is easily captured prematurely in the process of point cloud clustering, the improved particle swarm optimization algorithm, the social particle swarm optimization algorithm, is used to initialize the population and set a different follow threshold for each particle. To maintain individual diversity in the population, deepen the degree of global search for the population, and avoid falling into a local minimum; then, set the current optimal position of each particle in the population and the optimal location of the initial population, The position of free particles and the velocity and position of following particles are updated. Finally, fuzzy C-means clustering algorithm is used to solve the membership matrix, to determine the fitness function, and to update the optimal position of all particles. The location superiority of particle and population is judged, and the accurate cluster center is obtained, and the region division of point cloud data is realized. The point cloud model with inconsistent surface complexity is taken as an example to verify the algorithm. The feasibility of SPSO-FCM clustering algorithm is discussed and compared with FCM clustering algorithm and genetic FCM clustering algorithm. The experimental results show that the SPSOFCM clustering algorithm is faster than the other two algorithms in convergence speed, less iteration times, accurate clustering, clear boundary region segmentation, especially when the mechanical parts point cloud data with complex profile and more point cloud data are segmented. Better segmentation results can be obtained.
【作者单位】：南昌大学机电工程学院;赤峰学院建筑与机械工程学院;
【基金】：国家自然科学基金资助项目(No.51065021,No.51365037)
【分类号】：TP18;TP311.13

【相似文献】