KIVAOU - Analyse de séquences vidéo
The KIVAOU project aims at developing innovative tools dedicated to two main goals: - Mobile facial identification and biometric indexation tool (suitcase), with real-time video analysis - Platform for offline multiple video analysis of an event. These video are recorded during a particular event, and then collected for investigation purpose. After synchronisation between different sources, an analysis is performed in order to extract individual signatures, and to compute person's trajectories. The goal is to enable or to facilitate an offline analysis, for investigation purpose. The KIVAOU project proposes innovative approaches based on the combination of components already mastered by each partner: biometrics, image analysis, video analysis, tracking, synchronization, users GUI needs, etc. It includes a test phase with end-users.

– Face recognition and face tracking
– Improvement of the face quality by using surresolution.
– Selection of the best in term of quality, orientation and precision of series of images

– Pedestrian tracking
– Segmentation in region of interest, extraction of moving edges with morphological space-time gradient or background difference.
– Classification of the forms in classes: pedestrians, groups.
– Algorithm for components separation of the groups.
– Definition of signature based on the keypoints stored in KD-Tree for the multi camera re-identification

Number of scientific articles published: 11

Number of patents filed: 0

Number of product’s innovation: 2

software improvement:
– face recognition for Facing-IT
– pedestrian tracking for Evitech

Number of product’s innovation service: 0

Number of projected jobs created: 3

Number of jobs maintained: 2

2 full-time equivalents for the companies.

Number of related companies creation: 0