Specially made for a Raspberry Pi 4, see Q-engineering deep learning examples

A fully functional traffic counter with a camera working on a bare Rock 5C. Highlights:

• Stand alone.
• Lane selection.
• MQTT messages.
• JSON messages.
• Live web viewer.
• JSON settings.
• RTSP CCTV streaming.
• Debug screens.

To run the application, you have to:

• Optional: Code::Blocks installed. ($ sudo apt-get install codeblocks)

Start with the usual

$ sudo apt-get update 
$ sudo apt-get upgrade
$ sudo apt-get install curl libcurl4
$ sudo apt-get install cmake wget

Follow the Raspberry Pi 4 guide. Or:

$ sudo apt-get install libopencv-dev

$ sudo apt-get install libeigen3-dev

$ sudo apt-get install libgflags-dev

written by Niels Lohmann.

$ cd ~
$ git clone --depth=1 https://github.com/nlohmann/json.git
$ cd json
$ mkdir build
$ cd build
$ cmake ..
$ make -j4
$ sudo make install
$ sudo ldconfig

$ cd ~
$ git clone --depth=1 https://github.com/eclipse/paho.mqtt.c.git
$ cd paho.mqtt.c
$ mkdir build
$ cd build
$ cmake ..
$ make -j4
$ sudo make install
$ sudo ldconfig

$ sudo apt-get install mosquitto

Download the software.

$ git clone https://github.com/Qengineering/Traffic-Counter-Rock5C.git

Your folder must now look like this:

.
├── CMakeLists.txt
├── config.json
├── include
│   ├── BYTETracker.h
│   ├── dataType.h
│   ├── General.h
│   ├── kalmanFilter.h
│   ├── lapjv.h
│   ├── MJPG_sender.h
│   ├── MJPGthread.h
│   ├── MQTT.h
│   ├── Numbers.h
│   ├── postprocess.h
│   ├── STrack.h
│   ├── TChannel.h
│   └── Tjson.h
├── models
│   ├── yolov5m.rknn
│   ├── yolov5n.rknn
│   ├── yolov5s_relu.rknn
│   └── yolov5s.rknn
├── src
│   ├── BYTETracker.cpp
│   ├── kalmanFilter.cpp
│   ├── lapjv.cpp
│   ├── main.cpp
│   ├── main_video.cpp
│   ├── MJPG_sender.cpp
│   ├── MJPGthread.cpp
│   ├── MQTT.cpp
│   ├── postprocess.cpp
│   ├── STrack.cpp
│   ├── TChannel.cpp
│   ├── Tjson.cpp
│   └── utils.cpp
├── Traffic.cbp
└── Traffic.mp4

You can use Code::Blocks.

• Load the project file *.cbp in Code::Blocks.
• Select Release, not Debug.
• Compile and run with F9.
• You can alter command line arguments with Project -> Set programs arguments...

Or use Cmake.

$ cd *MyDir*
$ mkdir build
$ cd build
$ cmake ..
$ make -j4

All important settings are stored in the config.json
You can alter these to your liking. Please note the use of commas after each line, except the last one.

{
    "VERSION": "1.0.0.0",

    "MQTT_ON": true,
    "MQTT_SERVER_example": "broker.hivemq.com:1883",
    "MQTT_SERVER": "localhost:1883",
    "MQTT_CLIENT_ID": "Arrow",
    "MQTT_TOPIC": "traffic",
    "DEVICE_NAME": "highway 12",
    "ANNOTATE": true,

    "STREAM_example1": "rtsp:https://admin:[email protected]:554/stream1",
    "STREAM_example2": "RaspiCam",
    "STREAM": "Traffic.mp4",

    "BORDER_X1": 10,
    "BORDER_Y1": 300,
    "BORDER_X2": 450,
    "BORDER_Y2": 300,

    "JSON_PORT": 8070,
    "MJPEG_PORT": 8090,
    "MJPEG_WIDTH": 640,
    "MJPEG_HEIGHT": 480,

    "MESSAGE_TIME": 2,

    "PARAM_MODEL": "./models/yolo-fastestv2-opt.param",
    "BIN_MODEL": "./models/yolo-fastestv2-opt.bin"
}

You can use debug mode to find the optimal position for the borderline.
To enable debug mode, start the app with the --debug flag set to true:

./Traffic --debug=true

Alternatively, you can modify the command line argument in Code::Blocks by navigating to Project -> Set programs arguments...
In debug mode, you’ll see the tail of each vehicle. When a vehicle’s tail crosses the imaginary borderline, it is added to the count.
At this point, the bounding box is highlighted, which helps in identifying any missed vehicles.

You can receive MQTT messages locally at localhost:1883, the default setting. Messages are printed to the terminal.
When connected to the internet, you can send MQTT messages to any broker you choose, such as broker.hivemq.com:1883.
The app only sends messages when the MQTT_ON setting is set to true. The refresh rate, in seconds, is determined by MESSAGE_TIME.
At midnight, all cumulative counts are reset.

You can also follow the messages in a web browser. To do so, enter the port number after the IP address of your Raspberry Pi.

If your Raspberry Pi is connected to the internet, you can view live footage in a browser.
Simply combine the Raspberry Pi’s IP address with the MJPEG_PORT number specified in the settings to access the camera feed.