从大量含有噪声的3维点云数据中提取实物的边界特征,在以计算机视觉为基础的数字化曲面重建过程中有非常重要的意义。为提高重建精度,需要首先对大量原始散乱数据进行除噪及精简处理,但常规的数据处理方法由于没有区分噪声和特征点,因而使重建精度大大降低。为了准确的进行轮廓特征提取,提出了一种基于小波变换的激光测量扫描表面轮廓特征提取算法,并通过严格的理论推导,构造了一种类似m exh小波的小波基用来对两种边界特征点进行检测。多次实验结果显示,该算法不仅有效地避免了噪声和冗余数据的干扰,较精确地定位到了边界特征点,而且通过重建原始数据,较准确地提取了3维实体的外形轮廓,同时也为实现冗余数据的精简提供了一种新思想。

ion It is meaningful to abstract edge feature point from a large amount of 3D point cloud in the reconstruction of curved surface.The 3D data we get must be processed firstly in order to avoid the distortion and deviation during the course of reconstruction.But the common methods for this problem are of little use because noise points and the feature edge have not been analyzed and distinguished.According to this,a new approach based on wavelet edge detection is presented.We select a kind of wavelet similar to mexh as a tool exclusively to detect these two kinds of edge feature points,The following experiments show that: the edge feature points are located accurately by ignoring the disturbance of noise and redundancy.Comparing with former data proceeding methods,this method is more accurate and overcomes the influence of noise simultaneously.