图像三维重构中基于混合算法的基础矩阵估计

Fundamental Matrix Estimation Based on Improved Hybrid Algorithm in Image 3D Reconstruction

归一化8点法是基础矩阵线性估计方法中最经典的一种方法。为解决归一化8点法对误匹配抵抗性较差的问题,首先以8对匹配点作为基础矩阵估计的最小子集,利用归一化8点法估计相应的基础矩阵;然后运用粒子群算法对基础矩阵群体进行优化,去掉由误匹配造成的误差较大的基础矩阵,提高算法的准确性,在这个过程中,针对粒子群算法容易陷入局部极小的缺点,将混沌特性、自适应惯性权重调整机制以及模拟退火算法引入粒子群算法中,提高算法的搜索能力。实际应用表明,该方法提高了基础矩阵估计的精度和计算效率。

The normalized eight points method is seen as the most classic fundamental matrix linear estimation method. In order to solve the problem that normalized eight points method can＇t solve the effects of mismatching, first of all eight matching points are viewed as the fundamental matrix estimation minimal subset and the normalized eight points method is used to estimate the corresponding fundamental matrix; Then the particle swarm algorithm is utilized to optimize a set of the fundamental matrix for getting rid of the error fundamental matrix caused by mismatching points and improving the algorithm accuracy. In this process, the chaos characteristics, the adaptive inertia weight adjustment mechanism and the simulated annealing algorithm are introduced into the particle swarm algorithm to improve the ability to the search of the algorithm, because particle swarm optimization algorithm is easy to fall into local minimum. By the experiment result, the precision and computational efficiency of fundamental matrix estimation are improved.