Abstract: Airport target detection on high resolution remote sensing image carries several disadvantages, such as： miss detection, over detection, low robustness and inefficiency of dealing with massive pictures. An algorithm which consists of four-stage procedures is proposed to solve these problems. It pretreats the image with fuzzy enhancement. Edge detection is performed on the pretreated image to capture the contour of the main objects in the image. After that, it deletes the short and bending lines, and extracts the parallel lines from the rest long straight lines using Hough transformation. It locates the point near the parallel lines as the seed point of the region grow algorithm. Multi-process MPI parallel programming is used to improve the capability of processing massive pictures. The experimental results indicate that the proposed method is highly robust and can capture the airport target accurately. Besides, it greatly improves the performance in handling massive images.

Key words: airport target extraction, edge detection, parallel processing

摘要： 针对高分辨遥感图像机场目标检测存在漏检和误检，鲁棒性低，且处理多幅图片速度慢的特点，提出一种基于批量化并行处理模式的遥感图像机场目标提取方法。利用模糊增强方法对图像进行预处理，利用基于像素梯度和方差信息的模糊边缘检测算法对图像进行边缘检测，从中筛选出长直线，利用Hough变换提取其中平行的直线作为机场跑道特征。在得到的特征点中选取种子点进行区域生长，从而提取出完整的机场目标。利用MPI多进程并行处理的编程方法来实现对多幅图片中机场目标的批量化并行快速提取。实验结果表明该算法具有很好的鲁棒性，能准确地检测并提取出完整的机场目标，能够大幅度提高程序处理多幅图片的速度。

关键词: 机场目标提取, 边缘检测, 并行处理