Target Classification and Recognition for High-Resolution Remote Sensing Images: Using the Parallel Cross-Model Neural Cognitive Computing Algorithm

Target classification and recognition (TCR) are important information-extraction techniques for high-resolution remote sensing images (HRIs). However, because methods with high accuracy usually have higher time complexity, the massive remote sensing image has brought great difficulties for realtime application. In this article, we propose a hybrid, heterogeneous parallel processing algorithm to improve the efficiency TCR based on the Crossmodel Neural Cognitive Computing (CNCC) algorithm [parallel TCR based on MapReduce of hierarchical CNCC (PTCR-C)]. Parallel programming technologies used in this article include Open Multiprocessing (OpenMP), Compute Unified Device Architecture (CUDA), and the message passing interface (MPI). Experiments on public remote sensing image data sets show that the PTCR-C algorithm can effectively improve the efficiency of the original TCR algorithm. The PTCR-C algorithm has better augmented ability and provides the reference for further real-time TCR of remote sensing applications.