The Smart-City sample implements aspects of smart city sensing, analytics and management features as follows:
- Camera Provisioning: Tag and calibrate cameras for installation locations, calibration parameters and other usage pattern information.
- Camera Discovery: Discover and register IP cameras on specified IP blocks. Registered cameras automatically participate into the analytics activities.
- Recording: Record and manage segmented camera footage for preview or review (at a later time) purpose.
- Analytics: Perform analytics on the live/recorded camera streams. Latency-sensitive analytics are performed on Edge while others are on cloud.
- Triggers and Alerts: Manage triggers on analytics data. Respond with actions on triggered alerts.
- Smart Upload and Archive: Transcode and upload only critical data to cloud for archival or further offline analysis.
- Stats: Calculate statistics for planning/monitoring purpose on analytical data.
- UI: Present above data to users/administrators/city planners.
The sample implements the Smart-City traffic and stadium scenarios. The traffic scenario measures vehicle/pedestrian flow for planning purpose. The stadium scenario focuses on different access control techniques, including entrance people counting, service-point queue counting, and stadium seating zone crowd counting.
The sample is powered by the following Open Visual Cloud software stacks:
-
Edge Low-latency Analytics:
- The GStreamer-based media analytics stack is used for object detection, people-counting, queue-counting and crowd-counting on camera streams. The software stack is optimized for Intel® Xeon® Scalable Processors and Intel VCAC-A.
-
Smart Upload with Transcoding:
- The FFmpeg-based media transcoding stack is used to transcode recorded content before uploading to cloud. The software stack is optimized for Intel Xeon Scalable Processors.
-
Time Zone: Check that the timezone setting of your host machine is correctly configured. Timezone is used during build. If you plan to run the sample on a cluster of machines managed by Docker Swarm or Kubernetes, please make sure to synchronize time among the manager/master node and worker nodes.
-
Build Tools: Install
cmakeandm4if they are not available on your system. -
Docker Engine:
- Install docker engine. Make sure you setup docker to run as a regular user.
- Setup Kubernetes and helm.
- Setup docker proxy as follows if you are behind a firewall:
sudo mkdir -p /etc/systemd/system/docker.service.d
printf "[Service]\nEnvironment=\"HTTPS_PROXY=$https_proxy\" \"NO_PROXY=$no_proxy\"\n" | sudo tee /etc/systemd/system/docker.service.d/proxy.conf
sudo systemctl daemon-reload
sudo systemctl restart docker Use the following commands to build the sample. By default, the sample builds to the traffic scenario. To enable the stadium scenario, use cmake -DSCENARIO=stadium ...
mkdir build
cd build
cmake ..
make Use the following commands to start/stop Kubernetes services:
make update # optional for private registry
make start_helm
make stop_helm
The command make update uploads the sample images to each worker node. If you prefer to use a private docker registry, configure the sample, `cmake -DREGISTRY=, to push images to the private registry during each build.
Launch your browser and browse to https://<hostname> for the sample UI.
<hostname>is the hostname of the manager/master node.- If you see a browser warning of self-signed certificate, please accept it to proceed to the sample UI.
The sample supposes dynamic office starting/stopping. It supports the IP cameras deployed in the gateway and pushed to office via gb28181. You can selectively start and stop any office, as follows:
cmake -DNOFFICES=2 ..
make
SCOPE=cloud make start_helm
SCOPE=office1 make start_helm
SCOPE=office2 make start_helm
...
SCOPE=office1 make stop_helm
...
SCOPE=office1 make start_helm
...
SCOPE=office1 make stop_helm
SCOPE=office2 make stop_helm
SCOPE=cloud make stop_helm
A camera gateway aggregates the camera streams and pushes the camera streams to the edge offices. This is requred if the link between the cameras and the edge offices is a 5G network. Camera gateway is for legacy IP cameras that do not support GB28181. IP cameras that support GB28181 can push streams to the edge office(s) without a camera gateway.
The sample supposes dynamic starting/stopping of each office service and camera gateway, as follows:
At the edge cloud/office cluster:
cmake -DNOFFICES=2 ..
make
SCOPE=cloud make start_helm
SCOPE=office1-svc make start_helm
SCOPE=office2-svc make start_helm
...
SCOPE=office1-svc make stop_helm
...
SCOPE=office1-svc make start_helm
...
SCOPE=office1-svc make stop_helm
SCOPE=office2-svc make stop_helm
SCOPE=cloud make stop_helm
At the camera gateway:
cmake -DNOFFICES=2 ..
make
SCOPE=office1-camera make start_helm
SCOPE=office2-camera make start_helm
...
SCOPE=office1-camera make stop_helm
SCOPE=office2-camera make stop_helm
The sample supports running Cloud and Edge in different Kubernetes clusters. The Cloud cluster hosts the web server, the cloud database, and the cloud storage. Multiple Edge clusters can be present, each of which hosts a local database, camera discovery, camera streaming, and analytics instances. The Edge and the Cloud communicate securely through a connector host.
# Start Cloud instances
cmake -DNOFFICES=2 ..
make
make tunnels
SCOPE=cloud CONNECTOR_CLOUD=<user>@<connect-host> make start_helm
# Start Edge instances
SCOPE=office1 CONNECTOR_CLOUD=<user>@<connect-host> make start_helm
SCOPE=office2 CONNECTOR_CLOUD=<user>@<connect-host> make start_helm
...
SCOPE=office1 CONNECTOR_CLOUD=<user>@<connect-host> make stop_helm
...
SCOPE=office1 CONNECTOR_CLOUD=<user>@<connect-host> make start_helm
...
# Clean up
SCOPE=office1 CONNECTOR_CLOUD=<user>@<connect-host> make stop_helm
SCOPE=office2 CONNECTOR_CLOUD=<user>@<connect-host> make stop_helm
SCOPE=cloud CONNECTOR_CLOUD=<user>@<connect-host> make stop_helm
Specify the following optional parameters for cloud or office database high-availability settings:
HA_CLOUD=3 make start_helm
...
HA_CLOUD=3 make stop_helm
HA_CLOUD=3 HA_OFFICE=3 make start_helm
...
HA_CLOUD=3 HA_OFFICE=3 make stop_helm
Each database instance requires about 2GB memory.