Improving Cashewnuts Production through Automated Grading Using Enhanced Convolutional Neural Networks and Advanced Image Processing

India produces around 15% of the world's cashew
nuts, making it one of the top producers in the world. Tamil Nadu,
Kerala, and Karnataka are coastal states that are the main sites
used for farming. Over a million people in rural areas are
employed by the cashew industry, which makes a substantial
economic contribution to India through the export of cashew
kernels and cashew nut shell liquid. Using cutting-edge image
processing and deep learning techniques, this research attempts to
create an automated cashew grading system that improves
grading efficiency and accuracy. Digital cameras were used for
pre-processing cashew nut photos. Discrete Wavelet Transform
(DWT), Discrete Fourier Transform (DFT), and Discrete Cosine
Transform (DCT) were three compression techniques that were
carefully coupled to guarantee efficient feature preservation
under various image sizes and lighting circumstances. The CNN's
capacity to discern minute visual variations between classes is
enhanced by this hybrid compression technique, which eliminates
duplication while maintaining crucial spatial and frequency
information. The visual quality was further enhanced by applying
image enhancement techniques like thresholding, contrast
stretching, and homomorphic filtering.The performance of a
Convolutional Neural Network (CNN) built on the ResNet-50
architecture was assessed using common accuracy metrics after it
was trained on distinct databases created from every processed
image set. According to the findings, the suggested hybrid
compression and enhancement pipeline greatly increased
classification accuracy, making it possible to create a dependable
and entirely automated cashew grading system.