Abstract: Since a wood-flour microscopic image usually contains complex edge, fuzzy boundary and so on, an edge detection method based on object extraction and modified mathematic morphological multi-scale operator in the color space of HSV is proposed. This method extracts objects for avoiding background noise interference, and gains better wood-flour edge based on traditional multi-scale morphological operator combined with the best direction detection without the weight choice of the multi-direction detection. The experimental results demonstrate that morphological operator is better than traditional canny operator. The modified operator proposed contributes to a distinct edge detection result, improving detection accuracy as well as the capability of edge positioning.

Key words: wood-flour, edge detection, HSV color space, object extraction, morphological multi-scale operator

摘要： 针对木粉显微图像边缘复杂、细节模糊等问题，提出一种基于HSV空间目标提取和改进数学形态学多尺度算子的边缘检测算法。基于HSV空间提取目标，排除背景噪声干扰；通过最佳方向检测改进传统多尺度形态学检测算子，避免了多方向检测的权重选择，以此获得较好的木粉边缘。实验结果表明：形态学检测算子优于传统canny检测算子；改进的形态学多尺度检测算子不但取得了较好的边缘检测效果，而且提高了边缘检测精度和定位能力。

关键词: 木粉, 边缘检测, HSV颜色空间, 目标提取, 形态学多尺度算子