基于遗传变异的鸟群图像分割算法

Bird swarm image segmentation algorithm based on genetic variation

为避免鸟群算法在求取阈值时陷入早熟和局部优化,提出一种基于遗传变异的鸟群图像分割算法。在鸟群算法每一次迭代后加入遗传算法中的选择与变异,使算法加强在当前最优解附近的局部搜索能力,有效避免陷入局部最优解;针对噪声污染提出了"定位-中值-高斯平滑"滤波器对噪声污染图像进行预处理;将遗传算法、双种群遗传的算法、鸟群算法、遗传变异鸟群算法对图像进行分割的结果进行对比,以最大间类方差法(otsu)作为适合度函数。实验结果表明,遗传变异鸟群算法在寻找最优的解时的准确性和稳定性优于其它算法,对噪声有较强的鲁棒性。

To avoid the bird swarm algorithm getting into premature and local optimization when obtaining the threshold value,a bird image swarm algorithm based on genetic mutation was proposed.After each iteration of the bird swarm algorithm,the selection and mutation in genetic algorithm were added.The algorithm not only enhanced the local search ability near the current optimal solution,but also effectively avoided falling into the local optimal solution.A Positioning-Median-Gaussian smoothing filter for noise pollution was proposed to preprocess the noise pollution images.The genetic algorithm,two-population genetic algorithm,the bird swarm algorithm and genetic variation bird swarm algorithm were used to segment the images.The maximum variogram(otsu)was used as the fitness function.Experimental results show that the accuracy and stability of the genetic variation bird swarm algorithm in finding the optimal solution are superior to other algorithms and robust to noise.