TARGET DETECTION FROM HIGH-RESOLUTION SATELLITE IMAGES

Hits: 2581
Research areas:
Year:
2017
Type of Publication:
Article
Keywords:
Target Detection, High-resolution Satellite Images, Localised Adaptive Segmentation, k-NN, SVM
Authors:
Shibumon Alampatta, Narayan Panigrahi Partha Pratim Das
Abstract:
We present an object-based image classification method to detect aircraft from high-resolution satellite images. The detection of all varieties of aircraft is a difficult problem due to the large intra-class variability of aircraft objects, the presence of complex foreground / background scenarios in the image and the large volume of data to be processed. Further as the resolution of data increases the intra-object homogeneity decreases. In the proposed approach we use localised processing and leverage object-level attributes for classification. Localised adaptive segmentation is proposed for segmenting probable aircraft objects from the image and then object classification is performed using k-Nearest Neighbours (k-NN) and Support Vector Machine (SVM). Three band (Red, Green and Blue) data having about 0.5m spatial resolution are used in the experiments. We achieve an accuracy of 81% and 93% using k-NN and SVM respectively.
Full text: 84.pdf [Bibtex]
You are here: Home ACRS ACRS Overview Proceedings target-detection-from-high-resolution-satellite-images