关键词: 反距离加权插值(IDW), 空间邻近度, 各向异性, 多参数优化, 粒子群算法

Abstract: The inverse distance weighting (IDW) is one of widely accepted methods employed for predicting an unknown spatial value using known values observed at a set of sample locations. Many factors including spatial proximity, sample size and distance decay affect the estimation of the method simultaneously. However, most of IDW-based methods do not consider the effect of spatial anisotropy. In addition, most of these methods cannot predict accurate interpolating values because they only involve a sole factor for optimization. To obtain accurate missing values and high-resolution spatial surface model, the paper proposes a novel multiple parameters synchronization optimization IDW algorithm which involves anisotropy. The proposed method simultaneously optimizes the parameters of anisotropy, neighbor size and distance decay to improve the accuracy of IDW interpolation by particle swarm optimization (PSO). Moreover, scaling and direction factor are introduced to capture the varying of distance in different direction,and a new fitness function is schemed via cross validation technique. Two different resolution datum are selected to validate the effectiveness of proposed method, and the experimental results demonstrate that our method significantly outperform the typical IDW method. Comparisons with the recently developed IDW-based method, i.e. CIDW(classical IDW), FIDW(four quadrant IDW), AIDW(adaptive-IDW) and KAIDW(K-nearest neighbor adaptive IDW), OK(ordinary Kriging), as well as AnisOK(anisotropic Ordinary Kriging) are also implemented, and the experiments show that our algorithm can achieve the best interpolation results in terms of reliable and accuracy.

Key words: inverse distance weighting(IDW), spatial proximity, anisotropy, multiple parameters optimization, Particle Swarm optimization