基于FAST角点检测的遥感图像复原及目标轮廓矢量化

Remote Sensing Image Restoration and Target Contour Vectorization Based on FAST Corner Detection

结合高分辨率遥感图像的特点,应用FAST角点检测技术,提出了感图像复原及目标轮廓矢量化处理方法。建立遥感图像退化数学模型,推导出遥感图像的退化与复原函数,以特征角点的检测结果为图像复原的起始扩散坐标,通过连续拟合的方式实现遥感图像的复原。以复原图像为基础,通过标识轮廓边界、轮廓曲率计算和轮廓角点提取3个步骤,实现目标轮廓角点的检测。结合设置的矢量化约束条件,按照指定顺序连接角点,得出目标轮廓矢量化处理结果。通过实验分析得出结论:设计的图像处理方法可以有效的压缩35.6%的占用空间,与传统图像处理方法相比复原相似度更高、矢量化处理结果更精密。

Based on the characteristics of high resolution remote sensing image and FAST corner detection technology,a method of image restoration and object contour vectorization is proposed.The regression model of remote sensing image is established,and the regression and restoration function of remote sensing image is deduced.Based on the restored image,the detection of the target contour point is realized through three steps:marking contour boundary,calculating contour curvature and extracting contour corner.Combined with the set vectorization constraints,the corner points are connected in the specified order to obtain the result of vectorization of the target contour.Through experimental analysis,the conclusion is drawn that the designed image processing method can effectively compress 35.6%occupied space,and has higher recovery similarity and more precise vectorization results than traditional image processing methods.