Background Subtraction In Urban Traffic Video Using Recursive Sigma-delta Mixture Model

Motion segmentation is a fundamental step in urban traffic surveillance systems, since it provides necessary information for further processing. Background subtraction techniques are widely used to identify foreground moving vehicles from static background scene. Conventional techniques utilize single background model or Gaussian mixture model, which involves either poor adaptation or high computation.The complexity of urban traffic scenarios lies in pose and orientation variations, slow or temporarily stopped vehicles and sudden illumination variations. To address these problems Sigma-Delta Mixture Model (SDMM)is proposed. Mixed distributions are updated dynamically based on matching and contribution in the two order temporal statistics. The constant amplification factor is replaced byweightedfactor to update the variance rate over its temporal activity. The proposed technique achieve robust and accurate performance,which improves adaptation capability with balanced sensitivity and reliability, moreover, integerlinear operations enables the real-time capability.