基于改进二维熵-量子遗传算法的图像多阈值变化检测方法

Improved Quantum Genetic Algorithm for Maximization of Two-dimensional Exponent Entropy and Application in SAR Change Detection

针对二维熵法阈值分割中精度和时间性能较差的问题,提出了基于改进二维熵-量子遗传算法的多阈值图像分割方法。定义了二维阈值量子染色体的编码方式,解决了传统遗传算法优化二维最大指数熵阈值过程中速度慢、多样性小的缺点;在产生阈值解时,提出了半随机策略来代替传统的完全随机策略,加快寻优速度;改进了量子门旋转角度方式,提出了一种新的自适应旋转角度的方法,提高了算法的精度和收敛速度。并进行了分割实验和SAR 图像变化检测实验。结果表明:该方法比基于一维熵的图像分割算法具有更高的抗噪性;其寻优速度较完全随机产生阈值解的量子遗传算法提高了 3 倍~5 倍;避免了算法发散或过早收敛。与其他基于阈值分割的变化检测算法相比,性能更好。

Given the low accuracy and the poor time performance for the images segmentation,an improved two -dimensional exponent entropic quantum genetic algorithm for image threshold segmentation is proposed. Firstly,due to the poor time performance and the low biodiversity of the conventional genetic algorithm for image segmentation,quantum chromosome coding method based on two -dimensional threshold is proposed. Secondly,half -random generating thresholds strategy is put forward to accelerate the searching process. Finally,a new adaptive angle rotation strategy is proposed to improve the segmentation accuracy and convergence rate. The experiments show that the proposed algorithm has better noise immunity,and its searching efficiency is 300 %~ 500 % faster than that using random generating strategy. The improved algorithm can also avoid the divergence and premature convergence. The proposed algorithm can provide an effective solution for SAR change detection.