改进蚁群算法及其在高光谱影像分类中的研究  

Improved ant colony optimization algorithm and its research on hyperspectral image classification

作　　者：王偲晗 万幼川[1] 王明威 高雄 

机构地区：[1]武汉大学遥感信息工程学院,武汉430079 [2]中国冶金地质总局地球物理勘查院,河北保定071051

出　　处：《计算机工程与应用》2018年第1期196-203,共8页Computer Engineering and Applications

基　　金：国家重点专项(No.2016YFC0600210);国家科技支撑计划项目(No.2014BAL05B07);测绘遥感信息工程国家重点实验室开放基金(No.13R04)

摘　　要：针对高光谱影像波段数目多,易造成维数灾难的问题,结合遗传算法提供的初始启发信息和蚁群算法寻优能力的优势,提出一种基于改进二进制蚁群算法的波段选择方法。该方法通过遗传算法寻优获取几组较优解,经过计算后作为二进制蚁群算法的初始启发式信息,利用二进制蚁群算法的全局搜索获取最优解;另一方面,为充分利用影像的光谱与空间信息,将波段组合的光谱特征与改进二进制蚁群算法选择的纹理特征融合进行分类,可以获得更高的分类精度。实验结果表明,改进二进制蚁群算法与遗传算法、蚁群算法、二进制蚁群算法相比全局搜索能力更强,且该方法分类精度达到95.63%。An Improved Binary Ant Colony Optimization(IBACO)algorithm is proposed, which combined with the initial heuristic information of Genetic Algorithm(GA)and the global optimization ability of Binary Ant Colony Optimization(BACO)algorithm to overcome the high dimensionality of hyperspectral image. In the proposed method, the results of GA are utilized as the initial heuristic information of BACO, and the ant path selection mechanism is improved to enhance the global optimization ability. On the other hand, texture feature is utilized to make full use of the spectral and spatial information, and the combination of spectral and texture can obtain higher classification accuracy. Experimental results illustrate that the search ability of IBACO is better than GA, ACO, BACO. Meanwhile, the classification accuracy by using the proposed technique has reached 95.63%. In all, the proposed method can effectively enhance the classification efficiency and classification accuracy.

关 键 词：高光谱影像分类 改进二进制蚁群算法 波段选择 光谱特征 纹理特征 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TP751[自动化与计算机技术—控制科学与工程]