摘要
本文基于多特征融合,提出了一种运动目标识别方法。首先通过对运动目标的分割,分析得到各个目标的面积大小、形状复杂度;然后运用模板匹配方法,求得目标的运动速度。对上述特征进行模糊建模,提出相应的模糊规则,并采用模糊神经网络对推理系统的各个参数进行优化,进而识别目标。把这种识别系统用于对道路的监控,从而有效地识别道路中的机动车辆、行人以及摩托车/自行车。仿真试验表明,这种系统具有较强的学习能力以及识别精度。
A moving target recognition method is proposed in this paper, which is based on multi-features fusion. Firstly, all moving objects are segmented out, so the area and the degree of shape complex can be gotten for every object. By using matching method, the velocity of every target can be determined. A fuzzy reasoning system is then established, in which above characters are fuzzed. A fuzzy-neural network is used to optimize those parameters of fuzzy system, and then these moving objects are classified correctly. This system is used for road surveillance to distinguish motorcar, walking man, motorcycle /bicycle accurately. The experiences show that this system has a good adaptive ability and high accuracy.
出处
《系统仿真学报》
CAS
CSCD
2004年第5期1081-1084,共4页
Journal of System Simulation
基金
国家973项目(2001CB309403)
关键词
多特征融合
模糊推理
神经网络
识别
multi-feature fusion
fuzzy inference
neural network
recognition