双种群粒子优化算法在非线性光谱解混中的应用

Bi Swarm Particle Swarm Optimization Algorithm and its Application to Nonlinear Unmixing

光谱解混是高光谱遥感图像分析中的核心技术。近年来,非线性光谱解混引起了广泛的关注,但由于其模型的复杂性,导致求解方法十分复杂,此外,非线性模型很多,适用于不同的场合,传统的数值求解方法需要逐个模型推导其梯度等搜索方向,灵活性较差,导致其难以扩展。引用了粒子群算法(Particle Swarm Optimization,PSO),提出了双种群粒子优化框架,以适应各种非线性光谱混合模型,并以广义双线性模型为例,设计了非线性光谱解混算法。在实验中,采用了模拟数据和真实高光谱图像对算法进行了检验,结果表明,该算法可获得较高的光谱解混精度,能够很好地解决非线性光谱解混问题。

Spectral unmixing is an important technique for hyperspectral remote sensing analysis,of which theissue of nonlinear unmixing has recently attracted increasing attention.However,the complexity of the nonlinearmixing models(nonLMMs)increases the difficulty of developing a nonlinear unmixing algorithm.Furthermore,theapplication of various types of nonLMMs for different situations is hard to realized,because traditional algorithmsneed to derive the search direction(e.g.gradient)by models.To solve this problem,the Particle Swarm Optimization(PSO)is introduced and a biswarm PSO framework is presented,in order to be easily applied to different nonLMMs.Based on this framework,this paper develops a nonlinear unmixing algorithm,with generalized bilinear model(GBM)as an example.In experiments,synthetic and real hyperspectral datasets are used to evaluate the algorithm.Results indicate that the proposed algorithm outperforms other traditional algorithms and is an excellent method foraddressing nonlinear unmixing problem.