BODY WEIGHT ESTIMATION ALGORITHM FROM 2 DIMENSIONAL IMAGES BASED ON DIGITAL IMAGE PROCESSING

• Main Author: Juleoriansyah Nksrsb, Xl
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/71306
• Institution: Institut Teknologi Bandung

Body weight is an anthropometric feature that every single individual have and can be used along with body height to calculate a person’s body mass index (BMI) which can be used as a parameter for obesity. Not only that, body weight can also be used in order to adjust the amount of medication given to a specific patient. Both of those thing shows that body weight plays an important role in human life. However, there are certain situations where a person’s body weight can’t be measured through the conventional means, which leads to said person’s body weight being unkown. Through this study, a body weight estimation algorithm from 2-D images based on digital image processing is proposed as an alternative way for body weight measurement. This study could be divided into multiple steps, which include data acquisition, image pre-processing, image processing, and mathematical modelling. The pre-processing step is consisted of two steps, an RGB-to-greyscale conversion and gaussian blur application. The image processing step is divided into image segmentation, contour mapping, pose estimation, and feature extraction. The modeling will be done through the use of linear regression with Elastic-Net regularizatin On the evaluation done to the extracted anthropometric feature size, a Mean Absolute Percentage Error (MAPE) of around 10% was shown. On the second evaluation done to the model under optimal image condition the results are as follows, a Mean Absolute Error (MAE) of 3.84, a Mean Squared Error (MSE) of 17.18, and Coefficient of Determination (R Squared) of 0.97.