以具有典型形状特征的操场为例,探讨了形状特征在人工目标自动识别中的应用。基于形状特征的人工目标识别,一般首先将图像分割为多个基元,然后依据人工目标有较规则形状的特点,以形状特征作为指标进行识别。由于高分辨率遥感图像细节信息丰富,这种方法所提取的目标轮廓往往不完整。本文提出了"图像-基元-目标-轮廓恢复"的识别模式,在目标识别后采用加入形状先验知识的主动轮廓模型对目标进行轮廓恢复。实验结果表明,这种方法可以有效地修正目标提取结果中的轮廓缺失。 更多还原
【Abstract】 In this paper,we propose a new method utilizing prior shape to recognize an artificial object in remotely sensed image of high spatial resolution.As the classic object-oriented method,we get image parcels via threshold segmentation,and extract the target by its peculiar shape feature.Further on,we add a post-processing process,called contour revision,to improve the contour accuracy of artificial objects.The contour revision process is accomplished by an active contour model incorporating with pr... 更多还原