基于多尺度自适应显著区域检测的舰船三维重建外点消除

Outlier Elimination for Warship 3D Reconstruction Based on Multi Scale Adaptive Saliency Detection

针对海杂波影响下对海拍摄舰船图像,在三维重建过程中存在的大量冗余外点的问题,给出了一种多尺度自适应显著区域检测方法提取舰船区域,滤除海面信息,以消除密集点云中的外点.该方法利用大尺度因子高斯滤波,定位出舰船的大体轮廓,利用图像亮度分量显著图开方值提取舰船水面倒影,利用基于目标尺度的自适应高斯滤波提取舰船精细轮廓,从而得出较精确的舰船区域检测结果,可有效消除三维重建过程中外点的影响.

For the warship images with sea clutter,there are lots of outliers.This paper proposes a multi scale adaptive saliency detection method to extract the warship,filter the sea surface,and clear the dense point cloud.The method uses large scale Gaussian filter to locate the warship’s outline uses the brightness weight parameter’s saliency map extraction value to extract the water reflection,and then extract the fine outline based on the adaptive Gaussian filter function with the target scale.Finally,the accurate warship area can be extracted by eliminating the influence of outliers in a 3 D-reconstruction.