摘要
应用立体匹配技术,通过像点获取景物的距离信息,使得三维立体再现成为可能。图像的立体匹配是立体视觉中的一个关键问题,也是典型的"病态"问题。为了提高匹配的抗干扰能力以及全局最优性,提出了将改进的遗传算法用于立体匹配。选择图像边缘作为匹配基元,根据遗传算法容易获得全局最优解的特点,将图像的特征点进行染色体编码并重新设计比较了评价函数及相应的遗传操作,使之适合于立体匹配,并将图像的Rank变换加入其中,以提高算法的抗噪性。实验表明,该方法具有可行性。
Using stereo matching technique and the image point distance information can make the three-dimensional representation possible.The stereo image matching is a key problem for stereo vision,and also it＇s a typical ＂pathological＂ problem.In order to improve the matching anti-interference ability and global optimal solution,a new approach to stereo matching using the improved genetic algorithm was proposed.Image edge is used as matching primitive.According to the characteristic that genetic algorithm can easily acquire entirely optimal answers,feature points in the image are encoded as chromosomes.The evaluation function and the corresponding genetic operators are redesigned to fit in the stereo matching.Rank transform of the image is used in order to improve the antinoise ability of the algorithm.Experimental results show that the proposed method is effective.
出处
《计算机技术与发展》
2011年第8期46-49,共4页
Computer Technology and Development
基金
国家863项目(2008AA10Z211)
校博士基金项目(08zx7101)
