Soft Sensor for Capacitance Tracking in Quality Assurance of Fruit Ripening Process

Abstract

The highlight of this research work is to discover an efficient capacitance tracking for generation of ethylene gas (C2H4 in ppm) used for ripening of fruits by employing soft sensor built using image processing and Artificial Neural Networks (ANN) algorithms. The proposed method relies on the statistical analysis of the color which denotes the various stages in ripening and in turn indicates the amount of ethylene gas required. The changes in color, texture, intensity variation, mean, variance and standard deviation extracted from the images are the features which enable the personnel to determine the amount of ethylene gas. The Feed Forward Neural Network (FFNN) is used for ethylene gas and the corresponding capacitance value estimation. This is made possible using Back Propagation Algorithm (BPA) for training the FFNN. As a part of image processing the intensity values in color images and its variation are tracked by dithering which is used as a unique feature input to train the FFNN. The major findings of the proposed method depends on the FFNN estimating the ethylene gas needed for ripening process in a feed forward fashion thereby providing the precision and recall values spontaneously for every instance. The improvement made in application side denotes that earlier a circuit with capacitance is used to generate ethylene gas for this purpose which is on other hand replaced by using a soft sensor. Nearly 51 images of banana are considered for training and testing respectively. Testing and confirmation result shows the required precision and recall values are in range of 80 to 89% and 100% respectively.