Paper Title
Spectral Classification in Detection of Foreign Objects Present in Food
Abstract
The main objective of the present work is to provide a new approach for image recognition using Artificial
Neural Networks using spectral classification. Since the different objects possess different spectral properties, collecting the
spectral features from each samples and analyzing it for the contaminants detection is the basic idea of the work. Feed
forward network is used for the weights calculation and back propagation is used for the error minimization in weights
calculated. Training is performed for all the compounds including contaminants. In the testing part compounds with ROI
which falls in predefined contaminant boundary which are considered to be contaminants and the execution is done in real
time.
Keywords— Artificial neural network, spectral properties, spectral features, feed forward network, back propogatio,weights
calculation, ROI, Real time.