An efficient framework for object detection and classification in remote sensing images based on BOW and unsupervised classification models [articol]

Abstract

Enthused by the current development of satellite and remote sensing images have attracted extensive attention. Nowadays, large number of research areas are focusing on developing applications It is one of the most significant challenges in real-world applications. The remote sensor collects data by detecting the energy that reflected from the earth in various location information and its store, retrieve, manage, display, and analyze all types of spatial data even though the accuracy is not satisfactory. This paper considers the problem of object detection and recognition as the main problem, and it is motivated to provide a better solution by designing and implementing an Efficient Framework for Object Detection and Classification (EFODC) on remote sensing images. The efficiency is improved by applying various image processing stages such as Image Acquisition and preprocessing, Image Enhancement, Object Detection, Bag-of-Words creation, and Training - Testing process. The bag-of-words method enables the user to maintain ground truth values for classifying the objects and improves the accuracy of classification. EFODC is experimented. The performance is evaluated by comparing with the state-of-the-art methods. Comparing with the existing approaches the proposed framework obtained 97.88% of precision and 97.47% of recall over 3000 images.