Abstract
In this thesis, we propose a real-time gait recognition system for identifying the car owner in the outdoor parking lot. First, we apply the background subtraction to extract human region and normalize it to the same size and center. Subsequently, the optical flow method is applied to calculate the person’s gait motion, and the histogram of motion vectors is applied to extract the gait feature. Two similarity measure methods are applied there for different propose. The PCA+MDA dimensional reduction is used when training data is available and the Chi-Square distance is applied in the other situation. After gait recognition is constructed, The real-time car security system is introduced which uses the gait feature to verify the car owner's identity. Finally, the CASIA database and the three outdoor video in the parking lot are used to test our gait recognition system's performance. The experimental results show that our system can achieve a good performance.