No abstract

Existing methods for pedestrian motion trajectory prediction are learning and predicting the trajectories in the 2D image space. In this work, we observe that it is much more efficient to learn and predict pedestrian trajectories in the 3D space since the human motion occurs in the 3D physical world and and their behavior patterns are better represented in the 3D space. To this end, we use a stereo camera system to detect and track the human pose with deep neural networks. During pose estimation, these twin deep neural networks satisfy the stereo consistence constraint. We adapt the existing SocialGAN method to perform pedestrian motion trajectory prediction from the 2D to the 3D space. Our extensive experimental results demonstrate that our proposed method significantly improves the pedestrian trajectory prediction performance, outperforming existing state-of-the-art methods.

This paper presents a novel approach of license plate location. The proposed algorithm involves the following three steps. First, the vertical edges of the vehicle image are extracted by Sobel operator. Second, HSV color space and integral image are employed to locate candidates in yellow license plates and non-yellow license plates. Finally, connected component analysis is to locate the region of license plate accurately. Experimental results on a large volume of natural-scene vehicle plate image sets, which are extracted from low-quality video sequences, demonstrate that our technique achieves a verification rate of around 95% on yellow license plates and 99% on non-yellow ones. The total time of processing one yellow image is less than 0.1s and the non-yellow one is less than 0.05s, meeting the requirements of real-time application.

No abstract

No abstract