HIGHWAYS, RAILWAYS, CITYSCAPES SEGMENTATION

Highway segmentation example

ROI (Region Of Interest) detection with AI and Deep Learning thanks to a better understanding of contexts with semantic segmentation.
This allows the real-time detections to only happen in the required ROI without interfering with others (e.g. only detect people and vehicles on the crossing path over the road).
Each color of the image segmentation corresponds to a context (e.g. road, vegetation, ballast, …).
The semantic segmentation is made without human interactions and can be executed at start or on time basis and is compatible with fixed or PTZ cameras.

3 domains of expertise are known at the moment : highways, railways and cityscapes.

The segmentation can be coupled with realtime detections on the PHOENIX AI products.

Use cases :

• highway cameras

• railways cameras

• smart city cameras

• petrol stations

• harbours

• police vehicles

• drones

• autonomous vehicles

• …
  And last but not least : detection of what is important to detect for you on top of the objects already detected in this bundle.

The PHOENIX AI product involved :

SM_ALL with the ability to supply one segmentation with the video.
SM_ALL is a video pass-through with IP connectivity and PoE in/out with smart AI and Deep Learning processing in it at up to 60FPS in 4K :
1 PoE input with 100Mbits
1 100 Mbps output
1 PoE Output with 100Mbits

Workflow :

RJ45 IP Camera with PoE > Camera Stream Out > SM_ALL with PoE : decode, segmentations, encode, statistics > SM_All Stream Out with detection + statistics (JSON, XML, …) ( > VMS, UI Software, …)

Resolutions / Bandwidth :

From 250 pixels width, up to 4Kp60
Up to 100Mbits/stream

Compatible with : RGB, NIR, Thermal cameras.