This paper presents the clustering of multiplane high-resolution orthoimagery and multispectral satellite images. Two well-known clustering techniques k-means and ISODATA are usually used for classification. K-means clustering is used in this paper for the classification. Since the clustering of satellite images of pixel dimension greater than 1000 × 1000 has increased execution time, hence it is considered for the parallelism. This paper depicts the data parallelism exhibited by different threads in cores of a processor in the legacy system using GPU by assigning the tasks among different threads independently. A framework of parallel computation is exhibited for clustering multiplane high-resolution orthoimagery satellite images and Landsat MSS datasets. A parallel block processing implementation for clustering has been exploited and tested specifically on CPU achieving an efficient speedup on multicore processor by varying with 2, 4, 8, and 12 threads with variation in number of clusters 2, 4, 8, and 12. Around 10 samples of MSS sensor and high-resolution multiplane orthoimagery satellite images are considered for clustering with the usage of MATLAB 2017a environment. Hardware resources are efficiently used from the results obtained in parallel approach resulting in time depletion compared to serial k-means clustering. This approach can be applied for processing remote sensing images as results are acceptable.
CITATION STYLE
Rashmi, C., & Hemantha Kumar, G. (2020). Multithreading Approach for Clustering of Multiplane Satellite Images. In Remote Sensing and Digital Image Processing (Vol. 24, pp. 25–47). Springer. https://doi.org/10.1007/978-3-030-24178-0_2
Mendeley helps you to discover research relevant for your work.