Traffic counter Raspberry Pi 4

Traffic counter with a camera on a bare Raspberry Pi 4.

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

---

Introduction.

A fully functional traffic counter with a camera working on a bare Raspberry Pi 4. Highlights:

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

---

Rpi5_Traffic.mp4

---

Dependencies.

To run the application, you have to:

• A Raspberry Pi 4 with a 64-bit Bullseye operating system.
  
• The Tencent ncnn framework installed. Install ncnn
  
• Optional: Code::Blocks installed. ($ sudo apt-get install codeblocks)

Installing the dependencies.

Start with the usual

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

Libcamera

$ sudo apt-get install libcamera-dev

OpenCV

Follow the Raspberry Pi 4 guide. Or:

$ sudo apt-get install libopencv-dev

Eigen3

$ sudo apt-get install libeigen3-dev

gflags

$ sudo apt-get install libgflags-dev

JSON for C++

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

paho.mqtt (MQTT client)

$ 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

Mosquitto (MQTT broker)

$ sudo apt-get install mosquitto

---

Installing the app.

Download the software.

$ git clone https://github.com/Qengineering/Traffic-Counter-RPi_64-bit.git

Your folder must now look like this:

.
├── CMakeLists.txt
├── config.json
├── include
│   ├── BYTETracker.h
│   ├── dataType.h
│   ├── General.h
│   ├── kalmanFilter.h
│   ├── lapjv.h
│   ├── lccv.hpp
│   ├── libcamera_app.hpp
│   ├── libcamera_app_options.hpp
│   ├── MJPG_sender.h
│   ├── MJPGthread.h
│   ├── MQTT.h
│   ├── Numbers.h
│   ├── STrack.h
│   ├── Tjson.h
│   ├── TChannel.h
│   └── yolo-fastestv2.h
├── LICENSE
├── models
│   ├── yolo-fastestv2-opt.bin
│   └── yolo-fastestv2-opt.param
├── README.md
├── src
│   ├── BYTETracker.cpp
│   ├── kalmanFilter.cpp
│   ├── lapjv.cpp
│   ├── lccv.cpp
│   ├── libcamera_app.cpp
│   ├── libcamera_app_options.cpp
│   ├── main.cpp
│   ├── MJPG_sender.cpp
│   ├── MJPGthread.cpp
│   ├── MQTT.cpp
│   ├── STrack.cpp
│   ├── Tjson.cpp
│   ├── TChannel.cpp
│   ├── utils.cpp
│   └── yolo-fastestv2.cpp
├── Traffic.cbp
└── Traffic.mp4

3 directories, 39 files

---

Running the app.

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

---

Settings.

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://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"
}

Global parameter	Comment
VERSION	Current version.
MQTT_ON	Enable MQTT messages. 'true-false'.
MQTT_SERVER	MQTT server. Default localhost:1883
MQTT_CLIENT_ID	MQTT client ID. Default Arrow
MQTT_TOPIC	MQTT topic. Default traffic
DEVICE_NAME	Name of the camera, used in the MQTT messages.
ANNOTATE	Show lines, boxes and numbers in live view. Default true
STREAM	The used input source. It can be a video or a RaspiCam, or an RTSP stream, like CCTV cameras.
BORDER_X1	Left X position of the imaginary borderline.
BORDER_Y1	Left Y position of the imaginary borderline.
BORDER_X2	Right X position of the imaginary borderline.
BORDER_Y2	Right Y position of the imaginary borderline.
JSON_PORT	The JSON message port number.
MJPEG_PORT	The thumbnail browser overview.
MJPEG_WIDTH	Thumbnail width
MJPEG_HEIGHT	Thumbnail height
MESSAGE_TIME	Define the interval between (MQTT) messages in seconds. Default 2.
PARAM_MODEL	Used nccn DNN model (parameters).
BIN_MODEL	Used nccn DNN model (weights).

---

---

Debug.

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.

Rpi5_Traffic_debug.mp4

---

MQTT messages.

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.

---

Preview.

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.

---

LCCV.

We use the Bullseye LCCV camera code, which is a lightweight camera solution that leaves most computing resources available for deep learning tasks.
If you prefer to use GStreamer, you can build the application with the CAMERA flag set to OFF:

$ cd *MyDir*
$ mkdir build
$ cd build
$ cmake -DCAMERA=OFF ..
$ make -j4

---

Final remark.

There is plenty of room for improvement. The most obvious upgrade would be adding more computing power. The network used is very lightweight. We’ve also published this application for the Rock 5C, an affordable ($60) device with an NPU. With this setup, you’ll immediately see the impact of using a more powerful DNN model.

---