一种基于概率Hough变换的遥感图像中线目标检测新方法

线状目标检测是遥感图像中人造目标识别的重要环节,其技术关键是如何有效地从背景中提取目标的长直边缘。Hough变换是从图像中检测直线的经典方法,但遥感影像的画幅大、背景复杂,标准的Hough变换会检测出大量短直线而影响对有用目标的检测效果。本文提出了一种改进的概率Hough变换方法,首先对原始遥感影像分块,利用Canny算子检测出每个子图像的边缘,再通过Hough变换检测出每个子图中的直线,并对其中的短直线聚类分组以排除干扰,最后再从剩下线段中提取并筛选出平行线以完成长直线目标检测。实验表明,这种方法对复杂背景下的遥感影像中的线状目标有良好的检测效果。

A Novel Linear Target Detection Method Based on Improved Probability Hough Transform in Remote Sensing Imagery

Linear target detection is one of the important courses in the artificial target recognition processing from the remote sensing imagery.The internal key technology is how to extract the long straight edges of the target from the background efficiently.Hough Transform is one of the classic methods for line detection.In view of the large scale and complex background in one remote sensing image,Standard Hough Transform (SHT) may generate too many short lines to distinguish the useful long linear targets.An improved method based on probabilistic Hough Transform (PHT) is proposed.Firstly,it divides the primary remote sensing image into sub-blocks and Canny operator is applied to detect edges inside each blocks.Then SHT is used to detect short lines and cluster them into groups to avoid most of the interfering lines.Finally,the parallel line pairs are extracted and filtered from the rest lines to confirm the long linear targets.Experiments show that the novel method can detect the linear target efficiently from the complex background.