Development of a Gait Recognition System

In this era of innovation and technology, new things are being introduced daily in every field. Recently, a new biometric named gait is being researched a lot to develop it into biometric authentication and recognition system for intelligent surveillance. Gait is defined as a person’s way of walking and is unique to every person. Gait has various attractive properties making it superior to other biometrics. The aim of the project is to design a new gait recognition system based on one of the latest deep learning residuals convolutional neural network ResNet34. The results of applying ResNet34 to publicly available datasets CASIA-B clearly indicates that this model shows promise in this area. To develop a completely functional product using the proposed model, I first collected gait data of an individual by making them walk in front of camera several times, for this I took the online video as I was not able to go to the university and use pedestrian camera. Background subtraction is then applied to the video frames to extract silhouettes and various morphological operations to remove noise and shadows. Gait Energy Image is used as feature extraction technique and finally a pretrained ResNet34 model is used to perform identification at real-time for this I used the CASIA-B samples as collecting my own data was not possible. Comparing the obtained results with other state-of-the-art models, my model achieves highest accuracy in more than half of the cases. This indicates that ResNet34 can achieve ground-breaking results in gait recognition field. In future, using better pretreatment and feature extraction techniques, I strongly believe that ResNet34 can be the next state-of-the-art model in gait recognition.INTRODUCTION 8 CHAPTER 1. LITERATURE REVIEW 12 1.1 History 12 1.2 Previous Solutions 13 1.3 Gait Databases 14 1.4 Problem Definition 15 1.5 Methodology 16 1.5.1 Methodological Approach 16 1.5.2 Methods of Data Collection/Selection 16 1.5.3 Methods of Analysis 18 CHAPTER 2. DESIGN OF SYSTEM OF GAIT RECOGNITION 19 2.1 Mathematical Fundamentals of System Design 19 2.1.1 Data Acquisition 19 2.1.2 Pre-Processing 21 2.1.3 Feature Extraction 23 2.1.4 Classification 24 2.2 System Architecture 25 2.2.1 Architecture Design Approach 25 2.2.2 Architecture Design 26 2.1.3 Subsystem Architecture 26 2.3 Detailed System Design 28 2.3.1 CCTV Camera 28 2.3.2 Pedestrian Detection 29 CHAPTER 3 IMPLEMENTATION AND TESTING 35 3.1 Pre-Processing Algorithm Implementation 35 3.2 Gait Energy Image (GEI) 42 3.3 Graphical User Interface (GUI) 44 3.4 Class Diagram 47 3.5 Training And Testing Resnet 48 3.6 Implementation Of Resnet 49 3.7 Gallery Probe Testing 52 3.8 Results and Discussion 53 CHAPTER 4. LABOUR PROTECTION AND SAFETY IN EMERGENCY 55 4.1 Labour Protection 55 4.2 Emergency safety 59 CONCLUSION 62 REFERENCES 6