Machine Vision based Object Detection using Deep Learning Techniques

Abstract

The identification of items on the surface of the earth is widely known to be possible using hyperspectral images. To do classification and identify the various items on the image, the majority of classifiers just take into account spectral information. In this study, a neural network convolutional is used to classify the hyperspectral picture based on spectral and spatial properties (CNN). There are only a few areas in the hyperspectral picture. The multilayer perceptron aids in the categorization of visual characteristics into many classes while CNN builds the upper categorical level of strategic spectral and spatial aspects in each of the patch. The patch size of 13 × 13 is found to be sufficient to attain the best accuracy. Compared to other classifiers, CNN requires greater computing time for training and testing. In comparison to other classifiers, simulation findings indicate that CNN stores the hyperspectral picture with the best classification accuracy.