Human motion analysis in dark surrounding using line skeleton scalable model and vector angle technique

Abstract

Detecting human existence in video streams is a fundamental task in many video processing applications. It is considered as one of the most challenging task that has attracted researchers’ attentions in various fields. In this paper, a novel procedure is produced to model, analyze and recognize human motions in video streams. This project will introduced a method for human motion analysis in dark surrounding using digital image processing technique. In particular, this project aims to detect and recognize human motions performing walking and running motions. There are four major areas that are related in this project for human motion analysis: (1) developing human body structure based on human skeleton model, (2) tracking and data collecting human motion with side view, (3) recognizing human activities from image sequences, and (4) image processing technique using edge detection and vectors angle calculation. All algorithms are developed using MATLAB software. Segmentation is developed to reduce the amount of data and filters out the useless information. Two methods are proposed for angle calculation and activities classification. Results showed that angle between 153.76°-180° for method 1 and 49.64°-92.86° for method 2 is classified as walking while jogging is 95.17°-138.72° for method 1 and 22.62°-56.31° for method 2. These application has successfully manipulated complex movement which is walking and jogging in dark surrounding for a clear activity determination.