摘要
深度图像分割是基于特征关系图匹配的曲面物体识别中的关键技术之一,针对已有深度图像分割方法存在的问题,提出了一种基于二值形态学的水线区域增长算法对深度图像进行分割。首先用结构元素迭代腐蚀深度图像形成距离图像;然后根据距离图像计算极限腐蚀的集合,提取出目标的种子点;最后用条件粗化从种子点开始生长回到原尺寸但不使各区域相连,完成深度图像的分割。实验证明,算法的分割结果与人的主观视觉感知具有良好的一致性。
The segmentation for range image is one of the key technologies in curved object recognition based on attributed relation graph. A watershed region segmentation method based on binary morphology for range image is proposed. First a structuring element is used to erode the range image iteratively to form a distance image. Then the sets of final erode is computed according to the former distance image to extract the seeds points of the object. Finally the seeds points are grown to the original size by conditional thickening and each region isn＇ t joined respectively, thus the segmentation for a range image is completed. The experimental results are given to demonstrate that the segmentation result is well consisted with the subjective vision apperceive of a person.
出处
《光学技术》
CAS
CSCD
北大核心
2009年第3期326-329,共4页
Optical Technique
基金
国家基金资助项目(40671157/D0121)
贵州省教育厅自然科学基金青年项目(黔教科2007047)
关键词
深度图像
二值形态学
水线区域分割
极限腐蚀
条件粗化
range image
binary morphology
watershed region segmentation
ultimate erode
conditional thickening
