real time video stitching (360x220°) from multiple IP cameras and streaming to wooza -- 2

The system consists of : - acquisition : 6 to ~16/20 sources, that could be Mandatory Option 1 : 1080p or 2160p (or bigger!) standard H264 RTSP stream from an IP camera (typ. sony or axis) Desired Option 2 : local 2160p sources from a HDMI capture card Next year :-) : Option 3 : local directX sources (USB cameras - with SDK for direct integration) the software that has to be developped has no HMI of user interface. It should be stable for 24/7 operations. It shall be able to perform do the following processing : - realtime, medium latency (a few seconds is OK, target is less than 2 seconds if this is a simple achievable performance) video stitching/blending panorama reconstruction from the video sources - based on nVidia or Radeon VR/stitching APIs (or any working real time solution) - (best effort) capable of taking a standard Hugin .pto to transform images - if not, have a simple and straightforward way to define the transformation parameters - all geometrical transformation are static, so they will be the same during the processing - dynamically, we shall do brightness & white balance correction, and (best effort) vignetting (based on predfined parameters and simple (http GET or local sql) availability of lens aperture - if not, vignetting is processed, but static as for geometrical changes) in realtime. Conditions are outdoor landscapes, so changes could be done at less than 1 Hz - and 0.1 Hz should be OK) in order to have the best, panorama with no seams - adjust processing to the lowest stream frame rate of cameras (streams fps should be between 25/30 and 4fps, and generally very close one to the others, and will change very slowly during the day) - is based on INI file for configuration (ip camera and sources, image transitions, etc.) - should stream the reconstructed panorama - this stream shall be wooza 360 compatible - typ 2160p - H264 - RTSP - cropping and rescaling shall be configured (typ. choose 1080p, or 8k as output size) - (best effort) should be able to take each sound/audio stream from each camera and embed it in wooza directional stream format - a simple log file shall be generated (2 levels = debug, normal operation) plateform shall be windows (preferred - but not absolutely mandatory) deliverables : source code + binary + some simple docs a complete development plateform with all HW and dev. ressources is made available through distant connexion (VPN) delivery : march/april 2017 We look for long term collaboration as many evolutions are foreseen