Parallel Algorithm for Plastics Identification Using Wavelet Transformations

The paper presents realization of a fast parallel algorithm for training and classification of plastic specimens by using kNN classifier, and signs result from factors wavelet decomposition of the input signal. The chosen method of parallelization is based on a hybrid model, a combination of multithreaded processing and message passing interface (MPI). Multithreaded processing is done using the standard directives OpenMP, and for communication between processes is used message passing interface in the implementation MPICH2). Measurements are conducted for estimation of speed up and efficiency as a function of the core numbers, the number of parallel processed samples, the sample size and the different types of parallelization as a function of the number of samples.