论文部分内容阅读
文中提出一种新的基于数据空间正交基的多通道遥感图像混合像元分解算法.该算法通过在数据集中确定一个具有最大体积的单形体来搜索端元.与经典的单形体遥感图像端元提取算法如N-FINDR不同的是,本算法将原基于行列式的单形体体积计算,等价于一组正交基的模值乘积计算,从而大大提高了算法的计算效率;同时,由于顺序搜索概念的引入,确保了本算法总能获得相同的端元提取结果,而不同于N-FINDR的结果易受随机选取的初始值影响.此外,利用这组正交基,文中所提出的算法还可以同时完成端元个数的确定与丰度的估计两项工作.模拟与实际数据实验结果表明,文中提出的算法是一种快速准确的遥感图像混合像元分解算法.
In this paper, a novel hybrid pixel decomposition algorithm for multi-channel remote sensing images based on orthogonal basis of data space is proposed, which searches for end elements by defining a single-body with the largest volume in the data set.Compared with the classical single-body remote sensing image Meta-extraction algorithm such as N-FINDR difference is that the original algorithm based on the determinant of the volume of a single body is equivalent to a set of orthogonal basis modulus product, thereby greatly increasing the computational efficiency of the algorithm; the same time, due to The introduction of sequential search concept ensures that this algorithm can always obtain the same endmember extraction results, while the result that is different from N-FINDR is easily affected by the initial value of random selection.In addition, using this set of orthogonal basis, The algorithm can also determine the number of end elements and estimate the abundance at the same time.Experimental results show that the proposed algorithm is a fast and accurate hybrid pixel decomposition algorithm for remote sensing images.