Method of Using Neural Network Combined with D-S Theory to Carry Out HRR Target Recognition
-
摘要: 基于D-S(Dempster-Shafer)证据理论,比较和研究了相关数据和不相关数据的融合方法,分析了多传感器数据融合的算法:集中式融合算法和分布式融合算法.经过实验证明,执行分布式有反馈融合算法时的效果最好.然后利用该算法,提出了和线性内插神经网络相结合的识别方法.利用4种飞机的步进频率雷达的高分辨率一维距离像,将神经网络的识别结果作为证据分别送入传感器进行融合,进行识别研究.实验证明,与单纯利用神经网络的方法比较,目标的正确识别率得到了改善.Abstract: Based on the Shafer-Dempster reasoning theory, the method of coherent and non-coherent data fusion is studied. The algorithms of multi-sensor data fusion, which include center fusion and distributed fusion. It has been improved that the result of distributed recursive algorithms is better than others. A method of using LINN neural network combined with data fusion is developed to carry out the target recognition. Using high resolution range image of four kinds of target, the result of recognition is put into multi sensor, and is fused then. The correct recognition result is improved greatly by this method compared with using neural network only.
-
Key words:
- recognition /
- high resolution /
- neural networks /
- multi-sensor /
- data fusion
-
[1] 肖人彬,王 雪,费 奇. 相关证据合成研究[J]. 模式识别与人工智能研究, 1993,6 (3):227~234. [2]刘雷健,杨静宇.基于融合信息的物体识别[J]. 模式识别与人工智能, 1993,6(1):28~33. [3]Bogler P. Shafer-Dempster reasoning with applications to multi-sensor target classfication[J]. IEEE Transactions on Systems, Man, and Cybernetics, 1987,SMC-17(6):968~977. [4]李 宏,徐 晖,安 玮基于BP网络与D-S理论相结合的点目标状态下卫星及其伴诱饵的识别方法[J]. 国防科技大学学报,1997,19(2):53~58. -
点击查看大图
计量
- 文章访问数: 2955
- HTML全文浏览量: 284
- PDF下载量: 809
- 被引次数: 0
下载:
