Trajectory Pattern Detection by HMM Parameter Space Features and Eigenvector Clustering

We develop an object trajectory pattern learning method that has two significant advantages over past work. First, we represent trajectories in the HMM parameter space which overcomes the trajectory sampling problems of the existing methods. The proposed features are more expressive and enable detection of trajectory patterns that cannot be detected with the conventional trajectory representations reported so far. Second, we determine common trajectory paths by analyzing the optimal cluster number rather than using a predefined number of clusters. We compute affinity matrices and apply eigenvalue decomposition to find clusters. We prove that the number of clusters governs the number of eigenvectors used to span the feature similarity space. We are thus able to automatically compute the optimal number of patterns. We show that the proposed method accurately detects common paths for various camera setups.