Alarmbian is a cross platform DIY NVR at its core, but it can also be used to build your own smart cameras. The cool thing is you can use almost any board Armbian supports, an x86 mini PC, an old x86 based PC and probably Windows since the code is based on Java, FFMPEG, OpenCV and Deepstack. All testing is on Ubuntu 22.04 at this point.

This project is at the beginning stages, but the event system and database logging of events is in place. The plan is to add REST services in order to get data from the camera servers and have a client be able to use that data. I've already built one based on an older version of Alarmbian.

Low power and small footprint ODROID-XU4 handles six 4K/15 FPS H265+ streams.
Motion detection built in with the ability to add other types of realtime detection.
History image shows entire motion event in a single image.
Use Deepstack or SenseAI.

I'm leaving optimized install up to the user since there are so many ways to optimize FFMPEG, OpenCV, Deepstack, etc. This is truly a DIY system. I'll point you in the right direction hopefully.

Install FFMPEG

This is the center of the camera stream universe and where hardware acceleration can be very important. FFMPEG is used to stream off the high resolution video and can also be used to read the substream frames in OpenCV, etc. The nice thing is that if you copy the stream to disk there are no decode/encode steps to slow things down. Thus it's easy to stream multiple 4K cameras to disk using low end ARM based hardware. In my case I'm using H265+.

FFMPEG generic sudo apt install pkg-config then sudo apt install ffmpeg if hardware acceleration is in the package or you want generic version
FFMPEG from source
FFMPEG Intel QuickSync
FFMPEG CUDA build-ffmpeg warning: this will build static libs

Install Java

cd
Download Zulu JDK 17 for your platform using the .tar.gz
tar -xf zulu*
rm *.tar.gz
sudo mkdir -p /usr/lib/jvm
sudo mv zulu* /usr/lib/jvm/jdk17
sudo nano /etc/environment
Modify PATH and append :/usr/lib/jvm/jdk17/bin to the end
Add JAVA_HOME="/usr/lib/jvm/jdk17" on new line
Save file
Close shell and open a new one
java -version

Install Ant

cd
Download latest Maven bin.tar.gz
tar -xf apache-ant*
rm *.tar.gz
sudo mv apache-ant* /opt/ant
sudo nano /etc/environment
Modify PATH and append :/opt/ant/bin to the end
Save file
Close shell and open a new one
ant -version

Install OpenCV

There are several ways to install OpenCV such as my script or Video IO hardware acceleration, but select what is optimized for your platform. The main thing is to build the Java bindings. I typically include Python, Java and C/C++ just to cover all the bases. Python is good for quick and dirty prototyping. If you do run my install you only need to do the following because Java is installed:

cd
git clone --depth 1 https://github.com/sgjava/install-opencv.git
cd install-opencv/scripts
Edit config.sh and make changes as needed
./install-libjpeg-turbo.sh
./install-opencv.sh
Check *.log files

Install Supervisor

Supervisor will be used to start all the jobs up required for Alarmbian. We will place all logs in ~/logs.

cd
mkdir logs
sudo apt install supervisor

Install rtsp-simple-server

It makes sense to centralize camera streams and minimize traffic from the cameras. rtsp-simple-server makes this happen. Cameras like the Annke C800 only allow one connection to the substream, so a proxy is required. Substreams are used for analysis and live viewing, so more than one stream at a time is required.

cd
Download latest .tar.gz file
tar -xf rtsp-simple-server*
rm *.tar.gz
nano rtsp-simple-server.yml
protocols: [tcp]
Edit paths section to specify your substreams
./rtsp-simple-server
Test proxy on client
^C to exit Add Supervisor job
Reference configuration
sudo nano /etc/supervisor/conf.d/rtsp-simple-server.conf
sudo supervisorctl update
Test proxy on client
Check logs dir for issues

H2 database

H2 is used to store data from the Alarmbian application. Other data stores could be used as well with configuration and schema.sql changes,

cd
Download latest jar file (use Binary JAR link)
Example wget -O h2-2.1.214.jar https://search.maven.org/remotecontent?filepath=com/h2database/h2/2.1.214/h2-2.1.214.jar
java -cp h2*.jar org.h2.tools.Server -baseDir ~/ -tcp -web -ifNotExists -tcpAllowOthers
Start another shell on same machine
java -cp h2*.jar org.h2.tools.Shell -driver org.h2.Driver -url jdbc:h2:tcp://localhost/nio:test -user sa -password sa
quit
ls -al test.mv.db
^C to exit server shell
rm test.mv.db Add Supervisor job
Reference configuration
sudo nano /etc/supervisor/conf.d/h2.conf
sudo supervisorctl update
Test H2 client
Check logs dir for issues

Install Maven

cd
Download latest Maven bin.tar.gz
tar -xf apache-maven*
rm *.tar.gz
sudo mv apache-maven* /opt/maven
sudo nano /etc/environment
Modify PATH and append :/opt/maven/bin to the end
Add M2_HOME="/opt/maven" on new line
Save file
Close shell and open a new one
mvn -version

Download project

sudo apt install git
cd
git clone --depth 1 https://github.com/sgjava/alarmbian.git
cd alarmbian
Edit POM and change opencv as needed
Edit POM and change opencv.lib as needed
mvn initialize
mvn clean install
cp server/target/server-1.0.0-SNAPSHOT.jar ~/.
cd
sudo supervisorctl start h2
sudo supervisorctl start rtsp-simple-server
Use application.properties to make your cam1.properties configuration
java -Djava.library.path=/home/servadmin/opencv/build/lib -jar server-1.0.0-SNAPSHOT.jar --spring.config.location=cam1.properties
^C to exit app
If you see a SIGSEGV don't worry because ^C will not be used for shutdown. Add Supervisor job
Reference configuration
sudo nano /etc/supervisor/conf.d/cam1.conf
sudo supervisorctl update
Check logs dir for issues

Install Docker and run as non-root user

sudo apt install apt-transport-https ca-certificates curl software-properties-common
curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo gpg --dearmor -o /usr/share/keyrings/docker-archive-keyring.gpg
echo "deb [arch=$(dpkg --print-architecture) signed-by=/usr/share/keyrings/docker-archive-keyring.gpg] https://download.docker.com/linux/ubuntu $(lsb_release -cs) stable" | sudo tee /etc/apt/sources.list.d/docker.list > /dev/null
sudo apt update
sudo apt install docker-ce
sudo usermod -aG docker ${USER}
Close shell and open a new one
docker version

Deepstack

This is for optional Deepstack 2.0 support. This can be run from a different system if needed.There is no longer support for ARM32, but all the rest of the stack supports ARM32.Use ARM64 or x86_64 to run Deepstack. Make sure you use the correct version of docker run if you want hardware acceleration, etc. Below is the CPU based version.

Please note Deepstack is very CPU intensive without hardware acceleration. Trying to run multiple cameras and Deepstack on the same low end hardware will result in poor performance.

ARM64
- docker run --detach --restart unless-stopped -e VISION-DETECTION=True -e MODE=Medium -v localstorage:/datastore -p 5000:5000 --name deepstack deepquestai/deepstack:arm64
x86_64
- docker run --detach --restart unless-stopped -e VISION-DETECTION=True -e MODE=Medium -v localstorage:/datastore -p 5000:5000 --name deepstack deepquestai/deepstack
docker update --restart unless-stopped deepstack To stop
docker stop deepstack To remove volume
docker volume rm localstorage

Use case 1 ROCK64

Going the SBC route can be a bit more difficult than using an old X86 based desktop and Ubuntu. The main advantages are price, size and power consumption.

ROCK64 V3 SBC with 4G RAM
VIA Labs, Inc. VL711 SATA 6Gb/s bridge (USB 3 port)
KingDian 480GB 3D NAND 2.5 Inch SSD
Realtek Semiconductor Corp. RTL8153 Gigabit Ethernet Adapter (USB 2 port)
Raspberry Pi heat sink and 12V fan powered off ROCK64's 5V/GND pins

In this instance I'm testing the full stack including Deepstack object detection. one 4K camera with motion detection uses about 9% CPU. Once frames are sent to Deepstack it will spike to 80%. Detection is about 2 FPS since the code serializes calls.

sgjava / alarmbian