多通道信息融合的改进乘积规则

• 作 者: 马继涌 高文 陈熙霖
• 机 构: 哈尔滨工业大学计算机科学系
• 出 处: 电子学报
• 发表日期: 1999年08期
• 关键词: 多通道信息融合 乘积规则 分类

用传统的多通道信息融合乘积规则计算出的融合分值的动态范围比较小，融合效果不理想，本文在假定所有通道都能同时正常工作的情况下，提出多通道信息融合的改进乘积规则．基本思想是对原乘积规则中的概率密度和融合分值平滑，使得融合分值的动态范围增大．文中的两个实验表明，改进的乘积规则的识别率有明显提高．最后文中给出了在已知每个单一通道系统识别率的条件下，由理想融合方法集成出的系统识别率的上限计算公式．

The traditional multi-channel information fusion product rule has a smaller dynamic range and a less satisfactory fusion effect. This paper proposes an improved product rule of multi-channel information fusion under the condition that all the channels can work normally at the same time. The basic idea is to smooth the probability density and the fusion score in the original product rule so that the dynamic range of the fusion score is increased. Two experiments in this paper show that the recognition rate of the improved product rule is obviously improved. In the last part of this paper, the upper bound of the system identification rate integrated by the ideal fusion method is given under the condition of knowing the recognition rate of each single channel system.