Pedestrian Gait Parameter Measurement Using a Cascade of Edge and Corner Detection Technique

Abstract

ABSTRACT
Most modern surveillance systems currently rely upon Closed Circuit TV feeds monitoring system. This report documents a new approach towards automating pedestrian recognition within typical video footage. The proposed work shows how the gait parameters can be measured from the video footage using the mathematical theory of Geometry and computer vision and pattern recognition technique which can be further used for recognize individuals. When people walk normal to the viewing plane, as major gait information is available like the ankle, knee and hip joint points which are successfully extracted for indoor/outdoor data. We describe a new model-based feature extraction analysis is presented using Corner Detection technique that helps to measure the essential parameters which can be used for gait classification/recognition. Using corner detection algorithm can reveal the end points of the legs describing the feet and other parts. By analyzing the gait parameters extracted in footsteps for different frames, it is possible to recognize a pedestrian subject. In the preprocessing steps, the picture frames taken from video sequences are given as input to anny Edge detection algorithm which helps to detect edges of the image by extracting foreground from background also it reduces the noise using Gaussian filter. The output from edge detection is given as input to the Harris Corner Detection technique. Using the Harris Corner Detection technique the corners of each gait parameter can be obtained and later it will be measured. In the proposed work, we have used five parameters to measure. It is observed that when the camera is placed at 90 and 270 degrees, all the parameters used in the proposed work are clearly measurable. The efficiency of the model is tested on a variety of body position and stride parameters recovered in different viewing conditions on a database consisting of 20 subjects walking at both an angled and frontalparallel view with respect to the camera. The test results show clear values with a high level of confidence and in this way, we will establish a baseline analysis which can be deployed in recognition.

Keywords: Biometric, Canny Edge Detection, Harris Corner Detection, Gaussian Filter.