ABSTRACT

The phenomenon of mixed pixels in satellite imagery is very common. Most of the existing unmixing works are based on linear mixing model due to its simplicity. Fan model and generalized bilinear model consider the bilinear interaction between the pixels. But, in many cases, the pixels are supposed to have multiple interactions. In this work ‘Modified Bilinear model’(MBM) is utilized for the nonlinear unmixing process that considers the entire single, bilinear and multiple interactions into account. Even though many nonlinear unmixing models show improved results compared to linear, the nonlinear unmixing of linearly mixed pixels shows even worse results. The Particle Swarm Optimization(PSO) technique is used in many engineering optimization problems but none of them have attempted this technique for Nonlinearity detection. In this work, a new model, Particle Swarm Optmization and Modified Bilinear Model ‘PSO-MBM model’ is proposed to perform the nonlinearity detection and spectral unmixing of satellite images which generalizes the linear, bilinear and nonlinear mixing models. The performance of this detection strategy is evaluated by conducting experiments on both synthetic and real datasets. It is found that the proposed PSO-MBM model has shown better unmixing accuracy comparatively with an average Root Mean Square Error (RMSE) of 0.1411 and average Reconstruction Error (RE) of 0.0678.

摘要

卫星图像中像素混合的现象非常普遍。由于其简单性，大多数现有的拆解工作都基于线性混合模型。扇形模型和广义双线性模型考虑了像素之间的双线性相互作用。但是，在许多情况下，像素应该具有多种相互作用。在这项工作中，“改进的双线性模型”（MBM）用于非线性分解过程，该过程考虑了整个单个，双线性和多个相互作用。尽管许多非线性分解模型显示的结果都比线性模型好，但线性混合像素的非线性分解结果却更差。粒子群优化（PSO）技术被用于许多工程优化问题中，但是没有一个人尝试将这种技术用于非线性检测。在这项工作中，提出了一种新的粒子群优化模型和改进的双线性模型“ PSO-MBM模型”来进行卫星图像的非线性检测和频谱分解，从而推广了线性，双线性和非线性混合模型。通过对合成数据集和真实数据集进行实验，可以评估这种检测策略的性能。发现所提出的PSO-MBM模型具有相对更好的解混精度，平均均方根误差（RMSE）为0.1411，平均重建误差（RE）为0.0678。通过对合成数据集和真实数据集进行实验，可以评估这种检测策略的性能。发现所提出的PSO-MBM模型具有相对更好的解混精度，平均均方根误差（RMSE）为0.1411，平均重建误差（RE）为0.0678。通过对合成数据集和真实数据集进行实验，可以评估这种检测策略的性能。发现所提出的PSO-MBM模型具有相对更好的解混精度，平均均方根误差（RMSE）为0.1411，平均重建误差（RE）为0.0678。