Abstract: For data fusion algorithm based on improved fuzzy clustering, there are some disadvantages such as inaccurate fusion and low reliability of fusion. In order to solve the data fusion problem of multiple homogenous sensors measuring a certain feature of the same target without prior knowledge, this paper presents a data fusion algorithm based on adaptive fuzzy C-means clustering, which mainly applies adaptive fuzzy C-means clustering to data fusion. The algorithm firstly introduces adaptive coefficients to find cluster subsets of different shapes and sizes in improved fuzzy clustering, making the fusion result more accurate. Secondly, Kalman filtering principle and neural network prediction method based on multilayer perceptron are applied to the error covariance estimation, which improves the credibility of the fusion. The experimental results show that compared with the four classical data fusion algorithms, the algorithm has better results in the fusion of the four simulated data sets with the real data sets, and the advantages are particularly obvious in criterion functions and fusion errors.

Key words: fuzzy clustering, adaptive, multi-sensor, membership degree influence factor, data fusion

摘要： 针对基于改进模糊聚类的数据融合算法存在融合不精确、融合可信度较低等不足，为了解决多个同质传感器在无先验知识的情况下对同一个目标的某一特征进行测量的数据融合问题，提出了一种自适应模糊[C]均值聚类的数据融合算法，主要是把自适应模糊[C]均值聚类应用到数据融合中。该算法首先在改进的模糊聚类中通过引入自适应系数以发现不同形状和大小的聚类子集，使得融合结果更精确；其次将卡尔曼滤波原理和基于多层感知机的神经网络预测法应用到误差协方差估计中，提高了融合可信度。实验结果表明，与7种经典数据融合算法进行对比，该算法在4个模拟数据集与真实数据集上融合结果较好，特别在判别函数与融合误差方面优势更为明显。

关键词: 模糊聚类, 自适应, 多传感器, 隶属度影响因子, 数据融合