距离选通激光成像空间定位模糊C均值聚类分割法

针对距离选通激光成像对比度低、照度不均、图像模糊的特点，提出了一种基于空间定位的模糊C均值聚类方法（SPFCM）对目标进行分割。传统的模糊C均值聚类法存在以下缺点:一是需要预先获得目标分类数量，自适应性较差；二是对空间信息不敏感，导致目标轮廓不完整以及错误分类。针对上述缺陷，文中对传统算法进行了改进，引入了初定位的概念，首先利用最大类间方差法（Otsu 法）和数学形态学工具对子目标进行初步定位，再将其形心方位信息和灰度信息融合到聚类过程中，以较短的迭代过程实现不同目标的归类。实验结果证明基于空间定位的模糊C均值聚类法可以完整、有效地对距离选通激光图像进行提取分割，处理时间优于传统FCM。

Spatial positioning fuzzy C-means algorithm in segmentation of range-gated image

A fuzzy C-means algorithm based on spatial positioning was proposed to do the segmentation for range-gated image, which had the feature of low contrast, uneven illumination, and blurring. Object extraction is essential in image processing, providing the basic and necessary information for other methods. Traditional FCM algorithm needs the number of classes to cluster the data, which limits its adaptability. It also lacks in sensitivity of spatial information, resulting in misclassification as well as incomplete extraction of objects. For the above defects, the traditional algorithm was improved by pre-positioning. Firstly, median filter, Otsu method, and mathematical morphology method were applied to do the initial segmentation, obtaining the centroid and grayscale information of all targets, which took very short time. Then both of the centroid and grayscale information were used in clustering process, accomplishing the classification with fewer iterations and less time consuming than traditional FCM. Experiments indicate that the the Spatial Positioning FCM (SPFCM) is effective in segmentation of range-gated image, the targets can be extracted more completely and faster than traditional FCM algorithm. This new method can be applied to navigation, tracking and surveillance with range-gated imaging system.