A robust algorithm for extracting ROI from palm image taken by mobile phone

The palm image taken by the mobile phone has differences in angle, illumination, etc. The RobustPalmRoi algorithm reduces these differences through a series of handlers, thus extracting a stable ROI.

The following palms are taken by iPhoneX and have different angles. The corresponding ROI is below it.

Advantages of RobustPalmRoi algorithm:

1. Lower extraction requirements, just use your mobile phone.
2. Faster extraction speed, just tens of milliseconds.
3. Higher extraction stability, eliminate the differences such as angle, illumination, etc.

Some tips for capturing palms on mobile phones:

1. Five fingers open.
2. Take the whole palm.
3. Low illumination on the background.

Installation

Install library

There are some requirements if you want to install RobustPalmRoi library:

• OS Linux/Unix like.
• OpenCV installed.
• yaml-cpp installed. You should install it as a shared library. Step
  1. git clone https://github.com/jbeder/yaml-cpp.git
  2. cd yaml-cpp && mkdir -p build && cd build
  3. cmake -DBUILD_SHARED_LIBS=ON -DYAML_CPP_BUILD_TESTS=OFF -DYAML_CPP_BUILD_TOOLS=OFF -DYAML_CPP_BUILD_CONTRIB=OFF ..
  4. sudo make install

Steps:

1. git clone https://github.com/Leosocy/RobustPalmRoi.git
2. cd RobustPalmRoi && mkdir -p build && cd build
3. cmake .. && sudo make install

Install Python module

Please make sure the C++ library is installed successfully.

Python3.x required.

Steps:

1. cd pypackage
2. python setup.py install

Usage

C++

In your CMakeLists.txt, add these lines:

add_definitions(-std=c++11)
find_package(robust-palm-roi REQUIRED)
include_directories(${ROBUST_PALM_ROI_INCLUDE_DIR})
...
add_dependencies(${YOUR_PROJECT} ${ROBUST_PALM_ROI_LIBRARIES})
target_link_libraries(${YOUR_PROJECT} ${ROBUST_PALM_ROI_LIBRARIES})

Then you can use it in your source code like this:

#include <opencv2/highgui.hpp>
#include <robust-palm-roi/chain.h>

int main() {
  rpr::ChainBuilder builder;
  builder.SetConfigYaml(config_file_name);
  auto chain = builder.BuildAndInitChain();

  cv::Mat img = cv::imread(palm_image_file_name);
  cv::Mat roi;
  auto status = chain->Process(img, &roi);
  if (status.IsOk()) {
    cv::imshow("roi", roi);
  }

  return 0;
}

You can find the detail under samples/cpp_sample directory.

Python

from rpr import HandlerChain

chain = HandlerChain('your_config.yaml')

# Process image, and get bytes of roi, and save roi as image.
import io
with open('your_palm_image.png', 'rb') as palm:
    roi_bytes = chain.process_image(palm)
    roi = io.BytesIO(roi_bytes)
    with open('your_roi_image.png', 'wb') as roi_image:
        roi_image.write(roi.read())

You can find the detail under samples/py_sample directory.

Run samples

Make sure you have installed library and rpr.py before running examples.

Before you run cpp or python sample, please prepare some date follow this:

1. cd samples/palm_data
2. ./download.sh

Run cpp sample

1. cd cpp_sample
2. mkdir build; cd build; cmake ..; make -j
3. ./run_cpp_sample

Run python sample

1. cd py_sample
2. python run_sample.py

Contributing

Please see CONTRIBUTING.md