论文部分内容阅读
本文研究的重点是开发在胶片图象上进行多目标定位的数学基础和算法。一种具有坚实数学基础的多目标定位的探测器算法已经开发出来。该算法基于这一假设:即图象由分布成分混合组成,源于不同背景和图象中目标区域。采用最大似然估计和迭代群集算法,开发出一种从混合分布状态中分离出其主要成分的探测器算法。掌握了混合的成分,就可根据混合成分的分类和对目标形状、大小的识别进行目标定位。该算法已在一组图象上进行过测试(图象的选择基于预计的困难情况),证实了该算法在噪音和杂波背景中进行多目标定位的能力。
The focus of this paper is to develop mathematical foundations and algorithms for multi-target positioning on film images. A multi-target localization detector algorithm with a solid mathematical foundation has been developed. The algorithm is based on the assumption that the image consists of a mixture of distribution components originating from different backgrounds and target areas in the image. Using maximum likelihood estimation and iterative clustering algorithm, a detector algorithm was developed to separate its main components from the mixed distribution. Grasp the mixed ingredients, according to the classification of mixed ingredients and target shape, size of the target identification. The algorithm, which has been tested on a set of images (the choice of which image is based on the expected difficulty), confirms the ability of the algorithm to locate multiple targets in a noisy and cluttered background.