Alpha Trimmed Mean Filtering Based Flexible Discriminative Feature Extraction For Object Detection In Aerial Images

Abstract

Feature extraction is a difficult problem to be resolved because it requires a substantial amount of processing time during the detection process. A lot of research works are presented for extracting features from images using different techniques. However, feature extraction performance of conventional works was not adequate. Therefore, an Alpha Trimmed Mean Image Filtering based Flexible Discriminative Feature Extraction (ATMIF-FDFE) method is proposed in this paper for performing efficient object detection in aerial images. Initially, remotely sensed aerial image is considered as an input in ATMIF-FDFE method. After that, input image gets preprocessed by using Alpha Trimmed Mean Image Filtering in ATMIF-FDFE method. Alpha Trimmed Mean Image Filtering process is carried out in ATMIF-FDFE method to remove unwanted noise from the input image for obtaining the enhanced image or to extract the useful information. After preprocessing, feature extraction is carried out in ATMIF-FDFE method by using Flexible Discriminative Analysis with the preprocessed image. Flexible Discriminative Analysis in ATMIF-FDFE method extracts the image features like color, texture, shape, etc from preprocessed image for performing the object detection from the remotely sensed aerial image with higher accuracy and lesser time consumption. Experimental evaluation of ATMIF-FDFE method is carried out on factors such as feature extraction accuracy and feature extraction time and false positive rate with respect to aerial image size and number of aerial images.