A Distributed Hyperspectral Target Detection Algorithm Based on Background Reconstruction for Cloud Platforms

Target detection is an important research direction in the field of hyperspectral image interpretation. This paper proposes a Spark-based distributed and parallel implementation of a hyperspectral target detection method that integrates variational autoencoder with constrained energy minimization to detect the residual image. Our proposed method can significantly reduce the computation time in hyperspectral target detection by means of distributed computing and is scalable with the increasing size of hyperspectral data. In addition, we introduce an AUC score-based data pruning (ASDP) strategy as well as its distributed version that facilitates parallel processing in our proposed method. Specifically, we use a data pruning scoring criterion to evaluate the hyperspectral data and remove redundant data with insignificant impact on the final training accuracy. Experimental results on several real-world hyperspectral datasets demonstrate that the distributed and parallel implementation can reduce the total amount of computational data for target detection without accuracy degradation.