基于多蚁群协作的遥感图像特征提取方法

摘要

This paper presents a novel feature extraction method for remote sensing imagery based on the cooperation of multiple ant colonies. First, multiresolution expression of the input remote sensing imagery is created, and two different ant colonies are spread on different resolution images. The ant colony in the low-resolution image uses phase congruency as the inspiration information, whereas that in the high-resolution image uses gradient magnitude. The two ant colonies cooperate to detect features in the image by sharing the same pheromone matrix. Finally, the image features are extracted on the basis of the pheromone matrix threshold. Because a substantial amount of information in the input image is used as inspiration information of the ant colonies, the proposed method shows higher intelligence and acquires more complete and meaningful image features than those of other simple edge detectors.%该文提出一种基于多蚁群协作的遥感图像特征提取方法。首先建立遥感图像的多分辨率表示，并在不同分辨率图像上释放蚁群。低分辨率图像上的蚁群以相位一致性作为启发信息，高分辨率图像中的蚁群以梯度强度作为启发信息，两个蚁群通过共享的信息素矩阵实现协作，完成图像特征检测。最后根据信息素矩阵提取图像特征。由于该方法在不同分辨率上引入了多种启发信息，因而能够提取更加完整和有意义的图像特征，表现出更强的智能性。