多尺度分解及信号融合的3D模型重构

3D model reconstruction based on multiscale decomposition and signal fusion

提出了基于经验模式分解和信息融合的三维几何模型重构方法,包括3个步骤:首先球面参数化信号并将其映射到平面,进行均匀规则采样;再对平面信号进行限邻域经验模式分解和小波分解,利用得到的各个内蕴模式图层和小波系数进行融合;最后从图层信号得到不规则的原始映射信号,逆映射回三维几何模型信号。同时,提出了基于信噪比的三维几何信号融合性能评价指标,对重构模型进行验证。实验结果表明,该方法能够有效融合三维几何信号,基于限邻域经验模式分解的重构效果优于小波分解。

An improved method based on empirical mode decomposition and signal fusion to reconstruct the geometry model is presented.Firstly,different three-dimension signals are spherically parameterized,transformed to the plane signal and sampled regularly;then the translated signals are decomposed by EMD and waevlet decomposition.The decomposed signals are fused as an image using detail components called Intrinsic Mode Function（IMF） decomposed by EMD and wavelet coefficients by wavelet;At last,the two-dimension signal is inversely mapped to geometry signal.One indicator for three-dimension geometry signal fusion based on SNR is proposed to test the reconstructed model.The rusults show that this method can fuse geometry model efficiently and the effect by EMD is better than that of wavelet.