Tampere University has been working on objection detection and tracking in camera images and videos. This includes state-of-the-art deep learning methods that enable detection and tracking of humans as well as other relevant objects (vehicles, containers, etc.). The task requires the usage of high-end GPUs. The results from object detection and tracking will be fused with measurements from satellite signal receivers (GNSS receivers) and inertial measurement units integrated in smartphones to improve positioning and navigation accuracy in port environments, which is a crucial step towards the automatization of ports.
Recently, Nhan Nguyen from Tampere University, has been studying coordinate transformation methods (known as homography). Homography describes the relation between two images of the same planar surface in space. In port environments, for example, humans and objects can be tracked by several cameras. Two challenges are 1) to recognize if the same person or object is tracked by several cameras simultaneously and 2) to infer the person or object location on a two-dimensional map from the camera images. Homography describes the relation between two images of the same planar surface in space using a 3-by-3 matrix, which enables us to tackle the two challenges mentioned. Multiple methods for estimating the homography matrix, ranging from closed-form solutions to methods that minimize geometric distances, are studied.
Categories: General news
