SVM based concealed target quality monitoring system using millimeter wave radar

In this paper, a methodology has been proposed for industrial quality monitoring applications for non-invasive packaged goods quality estimation using MMW imaging. A MMW imaging radar has been designed at 60 GHz. Ceramic tiles were used and covered with the cardboard for concealed targets formation. A variety of experiments with different random crack tile and non-cracked full tile configurations were made. Wavelet feature based SVM classifier has been proposed for non-destructive quality inspection. Optimum SVM classifier has been modeled using gaussian kernel function along with fine tuning of kernel parameters and error constraints. On an independent test data set, appreciably low false alarm for cracked tiles and no false alarm for non-crack tiles has been successfully attained which validates the proposed model. Thereby, a robust wavelet feature based SVM classifier model has been developed for non-destructive quality estimation for industrial applications.