自适应对数双搜索人工蜂群策略异源图像配准

Adaptive Logarithmic Double Search Artificial Bee Colony Strategy for Heterologous Image Registration

人工蜂群算法存在收敛精度不足、容易陷入局部最优解的缺点。为此,提出一种基于自适应对数收敛双搜索策略的人工蜂群算法。该算法引入自适应系数和对数收敛系数改进雇佣蜂阶段和观察蜂阶段的搜索策略,防止算法陷入局部极值;同时引入双搜索策略,加快收敛速度并提高算法精度。然后在该改进人工蜂群算法的基础上提出一种自然场景下果园中苹果异源图像配准算法,采用SURF算法的非旋转因子与黑塞矩阵响应阈值两个参数作为改进后人工蜂群算法的初始蜂群二维向量,采用均方根误差函数作为适应度函数进行寻优搜索。在实验阶段引入樽海鞘群算法、蝙蝠算法和另一种改进人工蜂群算法与所提算法进行性能比较,实验结果表明,在全局进化次数固定的条件下,所提算法使人工蜂群算法的性能有了50%~90%的提升,且异源图像配准结果较传统SURF算法能够得到更多正确的特征点配对。

The artificial bee colony algorithm has the shortcomings of insufficient convergence accuracy and easy to fall into the local optimal solution.An artificial bee colony algorithm based on adaptive logarithmic convergence double search strategy is proposed.The adaptive coefficient and logarithmic convergence coefficient are introduced to improve the search strategy in the hiring bee stage and the observation bee stage to prevent the algorithm from falling into local extremum;at the same time,a double search strategy is introduced to speed up the convergence speed and improve the accuracy.This algorithm is applied and an improved artificial bee colony algorithm is proposed for the registration of heterologous images of apples in orchards in natural scenes.The algorithm uses the non-rotation factor of the SURF algorithm and the response threshold of the Hessian matrix as the initial two-dimensional vector of the improved artificial bee colony algorithm,and uses the root mean square error function as the fitness function for optimization search.In the experimental stage,the salps swarm algorithm,the bat algorithm and an improved artificial bee swarm algorithm are introduced to compare the performance.The experimental results show that the proposed algorithm improves the optimization performance of artificial bee colony algorithm by 50%~90%under the condition of fixed global evolution times;The results of heterologous image registration can get more correct feature point pairing than the traditional SURF algorithm.