The Smart City sample scans the specified IP block for any cameras. If found, they will be registered to the database and be invoked by analytic instances for streaming.
The sample implemented camera simulation to facilitate evalaution. Camera simulation requires that you have a dataset to simulate camera feeds. The build script includes a sample clip (to be downloaded after accepting the license terms.)
If you plan to use your own dataset, put the files under the simulation directory. Rename them to follow the convention of <dataset-name>_<scenario>.mp4. The dataset must be a set of MP4 files, encoded with H.264 (configuration: baseline, closed-GOP and no-B-frames) and AAC.
If unsure, it is recommended that you transcode your dataset with FFmpeg:
ffmpeg -i <source>.mp4 -c:v libx264 -profile:v baseline -x264-params keyint=30:bframes=0 -c:a aac -ss 0 <target>.mp4
The sample implements the ONVIF protocol to discover IP cameras on the specified IP range. Do the following to add IP cameras to the sample:
Assume an IP camera can be on and off during the streaming cycles, we need to define a way to uniquely identify an IP camera (so that we can associate any provisioning parameters to the camera.)
- If the IP address of an IP camera is statically assigned, then the pair of the assigned IP address and the OnVIF port(s) can be the camera unique identifier, as follows:
"ip": "192.168.1.114",
"port": [ 80, 9988 ]
where ip specifies the IP address, and port specifies the list of available ONVIF ports. The sample will use the ONVIF protocol to discover the camera RTSP URL.
- If the camera RTSP URL is fixed and available, the IP address, the port(s) and the RTSP URL can be the camera unique identifier, as follows:
"ip": "192.168.1.114",
"port": [ 80, 9988, 554 ],
"rtspurl": "rtsp://admin:[email protected]:554/h264/live.sdp"
where port includes the camera ONVIF and RTSP ports, which tells the sample not to use the ONVIF protocol on the specified IP and ports. Just accept the RTSP URL as is.
- If the IP address is assigned through DHCP, then we need additional information about the camera, such as the camera serial number or the MAC address. Either one can be the unique identifier. Use the following script to scan the network for IP cameras:
read && PORT_SCAN='-p T:80-65535 192.168.1.0/24' PASSCODE=$REPLY make discover
where PORT_SCAN specifies the port scanning commands (similar to the nmap commands, supporting -p and -Pn). The camera network is 192.168.1.0/24 and the port range is 80-65535. At the prompt, please enter the passcode of the IP camera as username:password. The output is similar to the following lines:
...
"device": {
"Manufacturer": "LOREX",
"Model": "LNB8973B",
"FirmwareVersion": "2.460.0001.5.R, Build Date 2017-08-11",
"SerialNumber": "ND021808019141",
"HardwareId": "1.00"
},
"networks": [
{
"Name": "eth0",
"HwAddress": "00:1f:54:2d:3d:1b",
"MTU": 1500
}
],
...
The camera identifier can then be specified as:
"device": {
"SerialNumber": "ND021808019141"
}
or
"networks": {
"HwAddress": "00:1f:54:2d:3d:1b"
}
Provisioning is a process of associating a set of application-specific parameters to a camera, for example, the GPS location, field of view, direction and positioning transformation. Developing provisioning tools is outside the sample scope. As a workaround, the sample stores the provisioning information in sensor-info.json and uses it to initialize the database. See provision.py for how to ingest provisioning data into the database should you need to develop such a provisioning tool.
The provisioning information is scenario specific. The traffic scenario parameters are as follows:
"address": "Tendem Way",
"location": {
"lat": 45.5415,
"lon": -122.9227
},
"algorithm": "object-detection",
"mnth": 75.0,
"alpha": 45.0,
"fovh": 90.0,
"fovv": 68.0,
"theta": 0.0,
The stadium scenario parameters are as follows:
"address": "South East Wing",
"location": {
"lat": 37.38369,
"lon": -121.95448
},
"algorithm": "crowd-counting",
"theta": 225.0,
"zones": [7],
"zonemap": [{
"zone": 7,
"polygon": [[0,0],[1023,0],[1023,1023],[0,1023],[0,0]]
}],
The provisioning information should be paired with the camera identifier. The following example associates the South East Wing provisioning parameters with an IP camera, whose serial number is ND021808019141,
"address": "South East Wing",
"location": {
"lat": 37.38369,
"lon": -121.95448
},
"algorithm": "crowd-counting",
"theta": 225.0,
"zones": [7],
"zonemap": [{
"zone": 7,
"polygon": [[0,0],[1023,0],[1023,1023],[0,1023],[0,0]]
}],
"passcode": "admin:admin"
"ip": "192.168.1.0/16",
"device": {
"SerialNumber": "ND021808019141"
},
or with statically allocated IP address(es) and the ONVIF ports.
"passcode": "admin:123456",
"ip": "192.168.1.114",
"port": [ 80, 9988 ]
where passcode specifies the camera username and password; ip can be either a specific IP address 192.168.1.114 or an IP address range such as 192.168.1.0/24; and port is a list of the camera ONVIF ports.
It is recommended that you always specify an IP address range to be associated with passcode.
Certain IP cameras may lock up for a period of time if fed with a wrong passcode.
Enable the IP camera discovering service by configuring DISCOVER_IP_CAMERA to true in Docker swarm/office.m4, Kubernetes/office.m4, or Kubernetes Helm/values.yaml.m4.
Due to Kubernetes limitation, if IP camera is enabled, the discovering service and the analytics instances will run on the host network for reliably receiving RTP streams.
Restart the sample. Your IP camera(s) should show up in the sample UI.
IP cameras that support GB28181 can push the streams to an SRS server, where the sample can take as camera inputs.
- Enable the SRS cluster server
In the office.m4:
define(`DISCOVER_IP_CAMERA',`false')dnl
define(`DISCOVER_RTMP',`true')dnl
define(`DISCOVER_SIMULATED_CAMERA',`false')dnl
- Configure sensor_info.json to uniquely identify the IP camera
"rtmpid": "34020000001320000003@34020000001320000004",
"rtmpuri": "rtmp://xxx.xxx.xxx.xxx:1935/live/34020000001320000003@34020000001320000004"
where rtmpid specifies the IP camera sip user id and channel id and rtmpuri is the URL link.