Classification and Detection of Skin Tones Using Big Data Machine Learning Algorithms Under Rapidly Varying Illuminating Conditions

摘要

Skin tone detection is a perceptual symptotic interspatial computational analysis in pixel segmentation extraction from an image to identify the skin components from non-skin background. In a country like India, skin tone detection is a very complex task due to presence of wide variety of skin tones, further; modeling an algorithm under different environmental conditions is even more complex. The main aim of this paper is to overcome the drawbacks of existing algorithms in acquiring accuracy. We used Big Data Analysis and Big Data Machine learning techniques on a complete set of data collected from more than 800 images of different persons/group under different illuminating conditions. In this paper we propose a real time skin tone detection algorithm under different illuminating conditions and compare its performance parameters like True Positive Rate (TPR)., False Positive Rate (FPR) and False Negative Rate (FNR)., accuracy, F-score, precision and recall with existing skin tone detection algorithms. The proposed algorithm outperformed the existing algorithms.