QR Code reading with HALCON for ARM on a Raspberry Pi 3

In this proof of concept, we want to show you how to easily integrate a QR code reader on an embedded device using MVTec HALCON and HDevEngine. In our example, we used a Raspberry Pi 3 Model B with the standard camera module v2.

Prerequisites

• Rasperry Pi 3 running Raspbian
• Raspberry Camera incl. a mounting device to attach it to the RPi case
• MVTec HALCON to create the machine vision application on your desktop machine (You can download a free trial version atwww.halcon.com/now)
• MVTec HALCON for ARM-based Platforms (www.mvtec.com/products/halcon-embedded/arm/)
• Video4Linux2 interface (use the driver that is included in the HALCON for ARM-based Platforms installer)
• Any ftp client or usb thumb drive for transfering the source files onto the RPi
• Makefile to build the application for the desired target platform (ARMv8)

Creating the machine vision part with HDevelop

In the final application, we want to use two HDevelop procedures: One that initializes the camera and the data code reader, and one that finds the data codes (in our case, QR codes) and returns a result message. Let's have a look how easily these procedures can be created using HDevelop.

First, we create the procedure init_acquisition.hdvp. Here, we open and configure the image acquisition device. The output variable AcqHandle can later be used to grab images of the camera.

open_framegrabber ('Video4Linux2', 4, 4, 0, 0, 0, 0, 'default', -1, 'gray', -1, 'default', 'default', 'default', -1, -1, AcqHandle)

Secondly, we create the handle for the data code reader. We specify that we want to detect QR codes.

create_data_code_2d_model ('QR Code', [], [], DataCodeHandle)

Thus, the output of this procedure are the handles AcqHandle and DataCodeHandle.

Then, we create the procedure find_qr_code.hdvp. Here, we grab an image using the AcqHandle.

grab_image (Image, AcqHandle)

Then, we find QR codes in this image using the DataCodeHandle.

find_data_code_2d (Image, SymbolXLDs, DataCodeHandle, [], [], ResultHandles, DecodedDataStrings)

The rest of the program simply assembles the output message, that we want to return in a shell.

get_system_time (MSecond, Second, Minute, Hour, Day, YDay, Month, Year) Date := Year + '-' + Month + '-' + Day $ '.2d' + ' ' + Hour $ '.2d' + ':' + Minute $ '.2d' + ':' + Second $ '.2d' if (DecodedDataStrings == [])      ResultMessage := 'No QR code found.' + ' (' + Date +  ')' else       ResultMessage := 'Decoded QR code(s): ' + DecodedDataStrings + ' (' + Date +  ')' endif

Incorporating these procedures in a C++ project using HDevEngine

In the C++ project, we use these HDevelop procedures directly, with HDevEngine. First, we define the location of the procedures.

std::string ext_proc_path = "./hdevelop_procedures"; HDevEngine().SetProcedurePath(ext_proc_path.c_str());

Then, we call the first procedure and get the output variables.

  HTuple  acq_handle;   HTuple  data_code_handle;   InitializeAcquisition(acq_handle, data_code_handle);

The function InitializeAcquisition looks like this:

  HDevProcedure     proc("init_acquisition"); HDevProcedureCall proc_call(proc); proc_call.Execute();   proc_call.GetOutputCtrlParamTuple("AcqHandle", &acq_handle);   proc_call.GetOutputCtrlParamTuple("DataCodeHandle", &data_code_handle);

Similarly, we call the second procedure in a loop to find the QR codes and return a result message.

while (keepRunning)     {          HString         result_message;         FindQRCodes(proc, acq_handle, data_code_handle, result_message);          printf(result_message);          printf("\n");     }

Building the application

When building and running the application, we need to set some parameters. Thus, we created a simple setup.env file.

export HALCONROOT=$(pwd)/halcon export HALCONARCH=armv7ahfneon-linux-gcc48 export PATH=$HALCONROOT/bin/$HALCONARCH:$PATH export LD_LIBRARY_PATH=$HALCONROOT/lib/$HALCONARCH

export NATIVE=yes export DISPLAY=192.168.0.1:0 export XLOCALEDIR=$(pwd)/arm-xilinx-linux-gnueabi/share/X11/locale export LD_LIBRARY_PATH=$(pwd)/arm-xilinx-linux-gnueabi/lib:$LD_LIBRARY_PATH

# To use the Video4Linux API, load the driver sudo modprobe bcm2835-v4l2

The result

When executing the application, the output looks something like this:

Note: If more QR codes are within the camera's field of view, the QR code reader uses the code with the best quality.

The Raspberry Pi 3 with the camera module