摘要
针对单传感器进行图像目标识别时识别率较低,判决阈值较为苛刻这一弱点,提出了一种小波矩和Dempster-Shafe(rD-S)证据理论相结合的多传感器信息融合图像目标识别算法。利用小波矩提取图像的平移、伸缩、旋转不变矩特征,BP神经网络获取待识别目标属性的基本概率分配,最后利用D-S证据理论将单传感器的识别结果进行决策级融合,完成图像目标的识别。仿真结果表明了该算法在图像目标识别中具有更高的精度和可靠性。
Aiming at the defections of low recognition rate and demanding decision threshold in single-sensor image target recognition system,this paper proposes a new multi-sensor information fusion system based on the wavelet moment and the D-S theory. The image＇s translation, scaling, and rotation invariant moment features are extracted using the wavelet moment, and the basic probability assignments of the targets are obtained through a BP neural network. Finally,the recognition results of the single- sensor are fused together using The simulation results show that the D-S evidence theory and the image target is recognized correctly. this method has a high accuracy and reliability.
出处
《火力与指挥控制》
CSCD
北大核心
2013年第10期38-41,共4页
Fire Control & Command Control
基金
国家自然科学基金(61171155)
陕西省自然科学基金(2012JM8010)
西北工业大学研究生创业种子基金资助项目(Z2012076)
关键词
小波矩
BP神经网络
基本概率分配
D—S证据理论
信息融合
图像目标识别
wavelet moment, BP neural network, basic probability assignment, D-S evidence theory,information fusion, image target recognition
作者简介
刘兵（1988-），男，山西朔州人，硕士生。主要研究方向：目标识别与信息融合。
