Use of firefly meta-heuristic algorithm in improving the accuracy of satellite image classification, Case study: Rafsanjan
Subject Areas : Spatial data infrastructures and standardisation
1 - Department of Civil Engineering, Technical and Vocational University (TVU), Tehran, Iran
Keywords: Artificial Neural Network, Classification, Support vector machine, Firefly,
Abstract :
Today, one of the methods of preparing land use and land cover is the use of satellite information and Geographic information systems methods. Also, satellite data is very important in preparing land use maps due to providing up-to-date and digital information and the possibility of image processing. On the other hand, in recent years, advanced classification methods such as artificial neural networks, fuzzy sets and intelligent networks have been widely used to classify satellite images. The aim of this study is to improve the classification accuracy of satellite imagery using a meta-heuristic algorithm. In this research, a new supervised learning classification method was introduced based on the firefly algorithm. For that purpose, two neural network classification algorithms and the support vector machines were used for land use map classification. Then the overall and kappa coefficient values for these two algorithms were calculated. Finally, the firefly algorithm was programmed in Python software, and after entering the images into the Python program, it was found that the parameters of the core function of the support vector method and the number of neurons in the neural network method were improved. The classification accuracy for the ASTER sensor with this algorithm increased by 1.6% in the neural network classification and by 3.8% in the support vector machine classification. Finally, the results showed that the firefly algorithm is a suitable classifier and can compete with other methods.
_||_