改进果蝇算法优化二维熵火灾图像分割方法

A flame image segmentation method based on improved fruit fly algorithm optimizing 2-D entropy

针对火灾图像分割中二维熵阈值分割方法计算量大、运行时间长、阈值选取不够准确导致分割精度不高等缺点,基于果蝇算法的寻优过程提出改进方法。结合果蝇算法和逻辑函数自适应调整果蝇算法的搜索距离,果蝇种群根据改进搜索距离在二维灰度空间内搜索图像分割阈值,通过优化和迭代果蝇的位置找到最佳分割阈值分割图像。将该方法和二维最大熵方法、遗传算法优化二维熵分割方法的结果进行对比。实验结果表明,改进果蝇算法优化二维熵阈值分割方法在分割效果、阈值选取的准确度和运行时间方面均优于遗传算法,提高了火灾图像的分割精度,且抗噪性能好,具有良好的有效性和实时性。

In flame image segmentation,in order to improve segmentation accuracy affected by inaccurate threshold selection,as well as to overcome the large amount of calculation and long computing time of the 2-D entropy threshold segmentation method,an improved method based on the optimization process of fruit fly algorithm is proposed.Fruit fly algorithm and logic function are combined to propose an adaptive step to improve the search distance of fruit fly algorithm,thus to search image segmentation threshold in 2-D gray space.By optimizing and iterating the position of the fruit fly,the optimal segmentation threshold to segment the image is found.The algorithm is compared with the 2-D maximum entropy method and the genetic algorithm optimizing 2-D entropy segmentation method.The segmentation results show that the segmentation method based on improved fruit fly algorithm and 2-D entropy is superior to genetic algorithm in segmentation effect,accuracy of threshold selection and running time.The algorithm has good anti-noise performance,good real-time performance and effectiveness.