GPU Acclerated Automated Feature Extraction from Satellite Images

摘要

The availability of large volumes of remote sensing data insists on higherdegree of automation in feature extraction, making it a need of the hour.Thehuge quantum of data that needs to be processed entails accelerated processingto be enabled.GPUs, which were originally designed to provide efficientvisualization, are being massively employed for computation intensive parallelprocessing environments. Image processing in general and hence automatedfeature extraction, is highly computation intensive, where performanceimprovements have a direct impact on societal needs. In this context, analgorithm has been formulated for automated feature extraction from apanchromatic or multispectral image based on image processing techniques. TwoLaplacian of Guassian (LoG) masks were applied on the image individuallyfollowed by detection of zero crossing points and extracting the pixels basedon their standard deviation with the surrounding pixels. The two extractedimages with different LoG masks were combined together which resulted in animage with the extracted features and edges. Finally the user is at liberty toapply the image smoothing step depending on the noise content in the extractedimage. The image is passed through a hybrid median filter to remove the saltand pepper noise from the image. This paper discusses the aforesaid algorithmfor automated feature extraction, necessity of deployment of GPUs for the same;system-level challenges and quantifies the benefits of integrating GPUs in suchenvironment. The results demonstrate that substantial enhancement inperformance margin can be achieved with the best utilization of GPU resourcesand an efficient parallelization strategy. Performance results in comparisonwith the conventional computing scenario have provided a speedup of 20x, onrealization of this parallelizing strategy.