The goal of this project is to track and decode multiple barcodes and show bounding boxes around them.
Currently this can detect a single barcode and display a bounding box around it in realtime.
This uses Android NDK Camera API to access the camera from the native (C++) layer and then draw the preview on a SurfaceView using the NDK ANativeWindow.
Using OpenCV to generate the bounding box with some image processing.
This NDK solution will give much better FPS rather than using OpenCV on the Java layer (with CameraBridgeViewBase.CvCameraViewListener2, can be found in "BarcodeTrackingJava" branch).
- Android API 24 (Nougat) or greater
- OpenCV-Android-SDK
- Clone or download the repo and open "NativeBarcodeTracker" project using Android Studio.
- Change the OpenCV Android SDK library path in CMakeLists.txt file to your downloaded
OpenCV-android-sdkpath.
To make compiling faster and to minimize the project size I only have it load and run ARM 64-bit (arm64-v8a). To add a different abi architecture, lets take ARM 32-bit (armeabi-v7a) as an example:
- In the build.gradle file add
, "-DANDROID_ABI=armeabi-v7a"in arguments and, armeabi-v7ain abi-filters. (You can also replace it witharm64-v8a) - Go into the downloaded OpenCV Android SDK and copy
OpenCV-android-sdk/sdk/native/libs/armeabi-v7a/libopencv_java3.sofile into jniLibs folder by creating a directory namedarmeabi-v7a.
CV_Main.cpp, CV_Main.h, Native_Camera.cpp, Native_Camera.h and Util.h were initially taken from sjfricke/OpenCV-NDK project. It helped a lot to understand the implementation of the NDK camera along with the native surface window.
Image_Reader.cpp and Image_Reader.h were taken from googlesamples/android-ndk/camera project. Google NDK camera sample is also a great source to understand the NDK camera implementation.
OpenCV barcode detection algorithm initially taken from https://www.pyimagesearch.com/2014/11/24/detecting-barcodes-images-python-opencv/ tutorial. It explains the algorithm step-by-step.