一种改进的PSO-MBCV算法的车底阴影分割  

Segmentation of bottom shadow of vehicle based on improved PSO-MBCV algorithm

作　　者：付梦印[1,2] 靳璐[1,2] 王美玲[1,2] 杨毅[1,2] 

机构地区：[1]北京理工大学自动化学院,北京100081 [2]复杂系统智能控制与决策教育部重点实验室,北京100081

出　　处：《系统工程与电子技术》2014年第7期1439-1445,共7页Systems Engineering and Electronics

基　　金：国家自然科学基金(91120003);国家自然科学青年基金(61105092);北京市自然科学基金(6101001)资助课题

摘　　要：针对当前车底阴影分割算法在复杂环境下鲁棒性较差以及最大类间方差(maximum between-class variance,MBCV)多阈值分割算法不能自动确定阈值个数的问题,提出利用峰值自适应方法自动确定MBCV多阈值分割算法中阈值个数;然后,以阈值的个数为粒子群优化算法(particle swarm optimization,PSO)中粒子的维数,提出了一种改进的PSO-MBCV算法的车底阴影分割。实验结果表明,该算法能有较低的误分类误差,能有效地分割出车底阴影。The current segmentation algorithms of bottom shadow of vehicle have poor robustness,meanwhile,the multilevel thresholds segmentation algorithm of maximum between-class variance（MBCV）method does not determine automatically the number of the thresholds.Therefore,firstly,the peak adaptive method based on image histogram is used to determine the number of thresholds;then,the number is considered as the particle dimension of the particle swarm optimization（PSO）algorithm,and the bottom shadow of vehicles based on an improved PSO-MBCV algorithm is proposed.The results show that the misclassification error（ME）can be deduced and the bottom shadow of vehicles can be effectively segmented.

关 键 词：最大类间方差 峰值自适应方法 粒子群优化算法 误分类误差 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]