Multi Camera Multiple Object Tracking [2017-2019]
For the H2020 5G PPP phase 2 project 5GCAR, VISCODA develops algorithms for localization and tracking of vehicles in image sequences using a multi camera setup.
For the trajectory planning in autonomous driving, the accurate localization of the vehicles is required. Accurate localizations of the ego-vehicle will be provided by the next generation of connected cars using 5G. Until all cars participate in the network, un-connected cars have to be considered as well. These cars are localized via static cameras positioned next to the road. The demonstrated scenario is a lane merge where a car on the accelaration lane merges into the traffic on the main lane.
To achieve high accuracy in the vehicle localization, the highly accurate calibration of the cameras is required. The camera based detection, localization, and tracking provide accurate vehicle desciptions which are used for computing trajectory recommendation for all participating vehicles.
- 5GCAR Pre-demonstration (Demo Video for MWC 2019)
- Explanation of 5GCAR Use Cases (Demo Video for MWC 2018)
- B. Cellarius, K. Cordes, T. Frye, S. Saur, J. Otterbach, M. Lefebvre, F. Gardes, J. Tiphène, M. Fallgren: "Use Case Representations of Connected and Automated Driving", European Conference on Networks and Communications (EuCNC), June 2019
Extended abstract (pdf)
- K. Cordes and H. Broszio: "Constrained Multi Camera Calibration for Lane Merge Observation", International Conference on Computer Vision Theory and Applications (VISAPP)), SciTePress, pp. 529-536, Feb. 2019
Paper preprint (pdf)
- "5GCAR Demonstration Guidelines", May 2018
- M. Fallgren, M. Dillinger, A. Servel, Z. Li, B. Villeforceix, T. Abbas, N. Brahmi, P. Cuer, T. Svensson, F. Sanchez, J. Alonso-Zarate, T. Mahmoodi, G. Vivier, M. Narroschke: "On the Fifth Generation Communication Automotive Research and Innovation Project 5GCAR - The Vehicular 5G PPP Phase 2 Project", European Conference on Networks and Communications (EuCNC), June 2017
Object Motion Estimation [ongoing]
For the estimation of motion models of moving objects in video, a motion segmentation technique is utilized. Motion segmentation is the task of classifying the feature trajectories in an image sequence to different motions. Hypergraph based approaches use a specific graph to incorporate higher order similarities for the estimation of motion clusters. They follow the concept of hypothesis generation and validation.
Our approach uses a simple but effective model for incorporating motion-coherent affinities. The hypotheses generated from the resulting hypergraph lead to a significant decrease of the segmentation error.
- K. Cordes, C. Ray’onaldo, H. Broszio: "Motion-Coherent Affinities for Hypergraph Based Motion Segmentation", International Conference on Computer Analysis of Images and Patterns, Springer LNCS volume 10424, pp. 121-132, August 2017
Paper preprint (pdf)
Presentation Slides (pdf)
Scene Reconstruction from Video [ongoing]
For the products VooCAT/CineCAT, various algorithms for camera calibration, tracking, structure from motion, and video segmentation are developed. The basis for the scene estimation is the usage of corresponding features which arise from a 3D point being mapped to diﬀerent camera image plane. By using a statistical error model which describes the errors in the position of the detected feature points, a Maximum Likelihood estimator can be formulated that simultaneously estimates the camera parameters and the 3D positions of feature points.
The camera path and the reconstructed scene are essential for the integration virtual objects into the video. The point cloud is used for 3D measurements, such as distances to or between different objects of the scene.
Resource Adaptive Sceneanalysis and -reconstruction (Resourcenadaptive Szenenanalyse und -rekonstruktion) [2019-2021]
Research project in cooperation with the Institut für Informationsverarbeitung (Leibniz Universität Hannover / LUH) targeting optimal 3D scene reconstruction accuracy and object detection reliability with resource limitations such as hardware specific computation power.