Application of ELM Model to the Motion Detection of Vehicles Under Moving Background

Artificial intelligence techniques can be applied to illegal parking detection. In the task, identifying the target vehicle’s motion helps decrease the false alarm. The optical flow algorithm is commonly used for motion detection. The algorithm recognizes the motion by telling whether the pixel has displacement between two consequence pictures. But the algorithm cannot identify the actual moving vehicle when facing pictures photographed by a moving camera since all pixels move. We propose a new motion detection system based on the combination of ELM and optical flow algorithm. This system can handle the pictures snapped by the moving camera. In this paper, a new dataset focused on this application is built. After achieving displacement information by optical flow algorithm, the ELM neural network is used to learn the feature between the changing background and target vehicle. The system has been tested on our dataset, evaluating the ELM outperforms other machine learning models.