基于改进人工蜂群正余弦优化的红外图像分割方法

针对红外图像背景复杂且分割难度较大等问题,提出了一种改进人工蜂群正余弦优化的红外图像阈值分割方法。首先是将二维Otsu函数作为蜂群算法的适应度函数；其次采用混沌对立的学习方法和差分进化的方法改进了初始化种群和蜜蜂搜索方程；然后利用改进的蜂群算法优化阈值,缩小阈值的搜索区域；最后利用正余弦法计算出全局最优解,该最优解即为分割的最佳阈值。实验结果表明:论文方法与Otsu法、k-means法、区域生长法以及分水岭法相比,图像目标区域分割的平均交并比为84.13,平均准确率为89.18,有效提高了红外图像的分割精度。

Infrared image segmentation method based on improvedartificial bee colony sine and cosine optimization

In view of the complex background of infrared image and the difficulty of segmentation,a threshold segmentation method of infrared image based on improved artificial bee colony sine and cosine optimization is proposed.Firstly,the two-dimensional Otsu function is used as the fitness function of the bee colony algorithm.Secondly,the chaotic opposition learning method and differential evolution method are used to improve the initial population and bee search equation.Then,the improved bee colony algorithm is used to optimize the threshold and reduce the search area of the threshold.Finally,the global optimal solution is calculated by the sine and cosine method,which is the optimal threshold for segmentation.The experimental results show that:compared with Otsu method,k-means method,region growing method and watershed method,the average intersection and union ratio of images target region segmentation is 84.13%,and the average accuracy rate is 89.18%,which effectively improves the segmentation accuracy of infrared image.