Abstract-Recently, a novel evolutionary global search strategy called Imperialist Competitive Algorithm (ICA) has proven its superior capabilities in optimization problems. This paper presents an application of ICA in automated clustering of remote sensing images. The proposed algorithm is basically a hierarchical two-phase process. At the first phase the original data set is decomposed into water bodies and land cover classes using near Infrared band's information. At the second phase, ICA has been applied to determine the number and centers of the land cover clusters using RGB band's information during an unsupervised clustering. The optimization is based on Fuzzy C-Means and an additional term for improving the accuracy of clustering. The method is applied on pan-sharpened IKONOS images of Tehran and 4 artificial data sets with different properties. Results obtained from applying the proposed method for both artificial data sets and RS image, indicate promising ability of this method in clustering data with unknown cluster number. Also the results show that the achieved overall accuracy can be available, better than 78% in comparison with other applied methods.

No abstract

A novel instrumental method for the light fastness assessment has been established. The accuracy of the visual method is affected by undesirable constraints, such as the different severity and the complications due to off-tone color change of the specimens. Thus, it was desired to develop an instrumental method of the light fastness assessment. In this regard, a neural network-based instrumental method of the light fastness assessment was developed. First, a proper light fastness panel was prepared, and then the visual light fastness assessments of experienced and inexperienced observers were collected. The accuracy and repeatability of the visual assessment results from these two groups of the observers were analyzed using different statistical criteria. The statistical analysis has shown that the mean of three trials of the inexperienced observers can be combined with the mean of the results obtained by experienced. Thus, all the results from the inexperienced and experienced observers were used to prepare a valid dataset for training a neural network. Different neural network structures trained with the prepared valid dataset. Among all of the implemented structures, the most accurate neural network structure is the one with one hidden layer and a 3-7-1 structure. The root mean square error and correlation coefficients of proposed 3-7-1 NN are 0.32 and 0.975, respectively, for the test sets. According to these results and the results from the comparison of the instrumental and visual assessment of the light fastness, it was concluded that the proposed neural network can be used for the instrumental light fastness assessment.