Gait Identification under Surveillance Environment based on Human Skeleton
Using ST-GCN network to classify the human gaits according to skeletons
Co-contributor github @lixirui142, Supervised by Ke Xu
Leading
In real-world scenarios, identifying specific individuals through surveillance can be a challenging task. Facial recognition can be disrupted by masks or other factors, and even without those disruptions, it can be difficult to obtain sufficient facial information from cameras at a distance. Therefore, we may consider gait recognition as an alternative method for identifying people. By analyzing the unique patterns in a person’s walking style, we can build a database of gait signatures to compare against individuals captured on surveillance footage.
General
In our study, we utilized OpenPose to extract skeleton data from videos and reconstructed and adjusted Spatio-Temporal Graph Convolutional Networks to identify gaits. To evaluate the effectiveness of our model, we conducted experiments using the CASIC-B gait database. If you would like to learn more about our approach and findings, please refer to our paper, which is available on Arxiv at the following link: https://arxiv.org/ftp/arxiv/papers/2111/2111.11720.pdf.
Structure
