基于L_1度量的Type-2熵模糊聚类红外图像分割

为了准确实现目标识别,从红外图像的特点出发,提出了将L_1空间度量的二型(Type-2)熵模糊聚类算法应用干红外图像分割.该算法首先通过L_1空间度量样本点与类别中最大最小值的距离,代替了传统聚类算法中样本点与聚类中心的聚类,然后根据熵模糊聚类算法获得上模糊隶属度和下模糊隶属度两个隶属度函数,并采用二型模糊融合得到隶属度函数,其中给出了一种权重加权降型算法.通过对实际的红外图像分割表明,这种算法能准确地实现红外图像分割,自适应性强,鲁棒性好,能够在复杂背景下获得较为理想的分割效果.

Infrared Image Segmentation Based on Type-2 Entropy Fuzzy Clustering Using L_1 Metric

In order to recognise objects accurately, an algorithm for infrared image segmentation based on type-2 entropy fuzzy clustering using L_1 metric is propose from the characteristic of infrared image. Firstly, the distances between sam-ple and cluster center are replaced by the distances between sample and the maximum or minimum value in classification using L_1 metric. Then the upper membership and lower membership are obtained through entropy fuzzy clustering. The membership is get by type-2 fuzzy fusion. A weighted type-reduction method is given. The experimental results show the infrared image can be segmented well by the proposed method. This method is robust and adaptive. The ideal results are obtained in complex background.