Feature Selection for Gabor Filter Based on Level Measurement using Non-Interacting Tanks Level Images

Abstract

Level measurement models using image-based classifiers (pixel-based datasets) are used for estimation purposes. Pre-processing is thought-provoking in proceeding out the image filter technique and classifying the level. The level scenario of a two non-interacting tank system plays a vital role in predicting the level. Level monitoring is done using the supervised learning method using instance-based filters (Gabor Filter) and selected base classifiers for level measurements. The main scope of this case study is to improve the level measurements from the two non-interacting tank scenarios using Artificial Intelligent algorithms. The suggested article includes the finest feature selection process to increase the accuracy performance attained by the designated classifiers like IBK Instance base classifier for different neighbourhood values and Tree category algorithm like Random Forest. The performance accuracy in level prediction obtained is 81.356%, the weighted Average of Receiver operator characteristics of (ROC) 0.931 are obtained by Random Forest Tree Category Classifier