TrafficXRoads: Guide
page 31
Reducing latency
For many traffic control scenarios, it is necessary for the system to have as low latency as
possible so the traffic events are communicated to the controller in the shortest time
possible. To do this you can do multiple optimizations that we will discuss further. It is
important to keep in mind that designing a system to meet low-latency goals will require
other tradeoffs.
Reducing latency in a camera
Latency introduced by the camera is caused by image processing and encoding latency.
Here are some tips how to reduce the latency in a camera as much as possible:
Resolution
: choose a lower resolution if possible. Higher resolutions imply more
data do be processed which may lead to higher latency.
Enhancements
: rotation, scaling, deinterlacing, noise reduction, and more can also
add latency. Reduce the image enhancements as much as possible, but be aware
of the effect on image quality, especially in night conditions.
Encoding
: H.265 has lower latency than H.264
Number of streams:
Limit the number of streams to the number you really need.
Each unique combination of video settings will require its own individual encoding
process, adding load to the camera processor, causing delay.
Frame rate:
Higher frame rate will reduce delays caused by buffers. For a stream
with 30 FPS, each frame will take 1/30 of second to capture. The expected latency
is
33
ms in buffers.
Audio:
Audio causes appreciable amount of latency. Remove the audio.
Bitrate:
Reduce the amount of data being generated and transferred. Find the
lowest bitrate in which the image quality is still sufficient for automatic processing.
On the camera side the process is largely an exposure, image enhancement, compressing
and packing. Roughly speaking, the time it takes to process one frame in the camera is
under 50 ms.
Reducing latency in the network
Limit the total data volume being sent through the network. Make sure that your network
has a good quality of service, and that all the hops within the network are configure to
suite your video demand. Connect the cameras directly to the unit if it is possible.
Reducing processing latency in TrafficXRoads
The FLOW framework enables configuration of different settings that lead to significant
reductions in processing latency. These include the reduction of smoothing buffer for the
tracker and increasing the frequency of trajectory evaluation. Both parameters are under
the settings of the analytics and can therefore be defined separately for each one.
