Abstract: In this study, Landsat satellite images of hammar marshes and surrounding district in (Dhi Qar) province in the South of Iraq are classified by Back Propagation Artificial Neural Network (BPANN) for years 1991, 2000, 2015 and 2017. Firstly, Principle Components Analysis (PCA) is applied on six bands of these satellite images using MATLAB programming and the information of all six bands concentrated in first three principle component and then blended to form integrated image. Then the integrated image is classified using proposed method (BPANN) method based on encoding elements. In this proposed method (BPANN) there are two paths are considered training and classification. The estimated coded descriptors are input to the training and classification phases of the ANN. It is intended to prove that the encoding capabilities can lead to improve the classification accuracy. The training is useful to indicate the basic information about image classes that represented by some specified statistical features while the classification uses the same features to produce the final classification results in terms of training results. Results evaluation is carried out for validation purpose. Then, quantitative and qualitative analysis is estimated to evaluate the performance of the proposed classification method.. The artificial neural network showed valued ability for classifying satellite images.
Ashraf S. Abdulla, Bushra Q. Al-Abudi and Mohammed S. Mahdi, 2019. Classification of Al-Hammar Marshes Satellite Images in Iraq using Artifical Neural Network Based on Coding Representation. Asian Journal of Information Technology, 18: 241-249.