An efficient parallelization method of Dempster–Shafer evidence theory based on CUDA