Lot of research in the field of human recognition is being carried out. Gait recognition is a relatively new approach, which is gaining momentum in biometrics. For the application of image processing in recognition this paper will propose a simple approach, namely the Silhouette Vector.
The Silhouette Vector is the distance of boundary points from the centroid of the silhouette, as it rotates 360 degrees. In addition to the silhouette vector, the silhouette image was divided into three equal parts vertically. Some statistical properties of these parts were computed, added to the silhouette vector and given to the PCA training system. Training was performed using silhouette and rectangular vectors for each subject. For testing the system, the nearest neighbour method was used as it is one of the simplest algorithms used for classification problems. The test subject is assigned to the class which is the minimum Euclidean distance from it. Inclusion of the additional features has improved the system performance greatly. Cumulative match score was used to analyse the system performance.
Table of Contents
1 Introduction
1.1 Biomertics
1.2 Gait as Biometrics
2 Related Literature Review
2.1 Model Free Approach
2.2 Model Based Approach
3 Gait Recognition System
3.1 Literature used for Implementations
3.2 Database Available
3.2.1 CASIA Gait Database
3.3 Pre-processing
3.4 Quality Factor Q (Our Minor Contribution)
3.5 Feature Extraction
3.5.1 Gait Cycle Measurement
3.5.2 Silhouette Vector Measurement
3.5.3 Rectangular Features(An Experiment!)
3.6 Principal Component Analysis
4 Recognition Scheme
4.1 Testing
4.1.1 Feature Extraction of Test Sequence
4.1.2 Template Forming
4.2 Template Matching
5 Performance Evaluation
5.1 Introduction
5.2 Biometrics Evaluation Terms
5.2.1 Receiver Operating Curve-ROC
5.2.2 Cumulative Match Characteristic Curve-CMC
5.3 Other Methods
6 Results and Conclusions
6.1 Results
6.2 Conclusions
7 Future Work
7.1 Future Work Plan
Objectives and Research Themes
The primary objective of this work is to develop an automated gait recognition system using video processing techniques. The research focuses on extracting unique behavioral signatures from a walking person's silhouette to identify individuals at a distance, addressing challenges related to noise and data redundancy.
- Gait cycle extraction and automation using a custom Quality Factor (Q).
- Development of Silhouette Vector-based feature extraction.
- Experimental investigation of "Rectangular Features" for performance enhancement.
- Implementation of Principal Component Analysis (PCA) for dimensionality reduction.
- Performance analysis using Cumulative Match Characteristic (CMC) curves.
Excerpt from the Book
3.4 Quality Factor Q (Our Minor Contribution)
After performing these pre-processing morphological operations we found that some of the image frames changed drastically due to noise, breaks and holes present in it. We have to do compromise with the quality of image frame to eliminate the complete noise from it. There are 75 frames for the first subject “fyc” and likewise there are 20 subjects. It will be very complicated task to select such frames those having low noise or ideally noise free. To overcome this problem we defined a quality factor ’Q’. We set the standard quality (Q) at 10, and every time when a closing operation is required, we reduced the value of ‘Q’ by one unit. Hence each frame is associated with a numerical value i.e. quality factor. This is used to calculate the quality of the cycle and the cycle with the highest quality value is used for feature extraction.At this step we calculated quality factor,width of bounding box which is applied to each frame along and the area of image.
Summary of Chapters
1 Introduction: Provides an overview of biometric technologies, distinguishing between physiological and behavioral traits, and defines gait recognition.
2 Related Literature Review: Surveys existing methodologies in gait recognition, categorizing them into model-free and model-based approaches.
3 Gait Recognition System: Details the system architecture, including database selection (CASIA), pre-processing, feature extraction methods (silhouette and rectangular), and PCA implementation.
4 Recognition Scheme: Describes the testing phase, feature extraction for test sequences, template forming, and the use of the Euclidean distance classifier.
5 Performance Evaluation: Explains standard biometric evaluation terms and methodologies, including ROC and CMC curves for system performance analysis.
6 Results and Conclusions: Presents experimental results, discusses the impact of feature selection, and summarizes the findings of the research.
7 Future Work: Outlines limitations of the current study and potential directions for future improvements, such as handling opposite walking directions and refining pre-processing.
Keywords
Gait Recognition, Biometrics, Silhouette Vector, Principal Component Analysis, CASIA Database, Image Processing, Feature Extraction, Rectangular Features, Quality Factor, Euclidean Distance, Behavioral Biometrics, Template Matching, Performance Evaluation, Cumulative Match Characteristic, Human Identification
Frequently Asked Questions
What is the primary goal of this research?
The research aims to implement a gait recognition system capable of identifying individuals based on their walking pattern from video data using silhouette extraction and statistical analysis.
What are the main thematic fields covered in this study?
The study covers biometric identification, computer vision-based human gait analysis, feature engineering (silhouette and rectangular), and classification techniques like PCA.
Which classification method is utilized in this system?
The study employs a Nearest Neighbor method using a Euclidean Distance Classifier to determine the identity of the test subject based on the similarity between templates.
What is the significance of the "Quality Factor Q"?
The Quality Factor Q is a custom metric introduced to automate the selection of high-quality gait cycles from the video data, helping to mitigate the impact of noise and frame degradation.
What does the main part of the document address?
The main part focuses on the technical implementation: extracting features, refining them via a quality factor, applying PCA for dimensionality reduction, and evaluating the performance of the resulting gait signatures.
Which specific features are used for recognition?
The system uses Silhouette Vectors (distance of boundary points from the centroid) and newly proposed "Rectangular Features" derived from dividing the silhouette into three parts.
How does the system handle view variations?
The research primarily utilizes side-view sequences from the CASIA dataset and notes that performance decreases when the subject is walking in the opposite direction compared to the training data.
What is the role of PCA in this work?
Principal Component Analysis is used to reduce the dimensionality of the high-volume feature space while retaining the most significant variations for efficient and effective template matching.
- Citation du texte
- Vijaya Ahire (Auteur), 2019, Application of image processing in recognition, Munich, GRIN Verlag, https://www.grin.com/document/462126
