Satellite Image Classification using Back Propagation Neural Network

Objectives: The objective of proposed method is to improve the accuracy and performance of an image. To apply the segmentation on satellite image using a median filter and the Scale Invariant Feature Transformation algorithm for feature extraction from an image. And to implement the Back Propagation Neural Network algorithm so as to reduce the mean square error rate, false acceptance and rejection rate and to improve the accuracy of an image. Methods/Statistical Analysis: The single step pre-processing of an image is done to make it suitable for segmentation. For segmentation, median filter is used and the Scale Invariant Feature Transformation (SIFT) algorithm is used for feature extraction. When features from the image are generated then the image is optimized using a Genetic Algorithm. After image optimization Back Propagation Neural Network is used to classify the image based on different parameters. Our proposed technique is a Genetic Algorithm for an image optimization and Back Propagation Neural Network (BPNN) for classification of the satellite image. Findings: The image feature extraction using Scale Invariant Feature Transformation (SIFT) method results in better feature extraction as it gives both the key distribution points saved in database and image. The mean square error using GA-BPNN is less than existing technique ABC-FCM which gives better performance. GA-BPNN technique gives more accuracy which is approximately 99.91 as compared to other methods. Application/Improvements: The proposed technique has been tested with the images of different resolution and the results obtained by BPNN are proven to be better than the ABC-FCM. The proposed method can be used for different types of images and also for medical images.