Dynamic system stochastic identification mixed with video processing: Validation on a real case

Nowadays, visual information is everyday more present, cameras are more and more fast, small and accurate. Improvements in computer vision enable to go further: to consider cameras as smart sensors. In this framework, this article propose to mix complex mechanical models with video processing in order to estimate invisible parameter and to enable the camera to forecast in the video, with a dynamic meaning. The concept is developed on particular cases of objects moving in a scene where the scenario is known. This application is a proof of concept on a case study of a 2D scenario of a ruler sliding on a table. The aim is to prove the feasibility on a real case with real time architecture.