200 likes | 331 Vues
This project focuses on real-time detection and reading of QR codes on Android devices, adhering to ISO/IEC standards. It addressed challenges such as various lighting conditions and perspective distortion, using an approach based on Fast Radial Symmetry for detecting points of interest. Key tools included MATLAB, Eclipse, and OpenCV. The project also involved outlier filtering and tuneable parameters for enhanced detection performance. Achievements included successful parameter determination and QR code reading across different sizes and conditions, culminating in a live demonstration.
E N D
Android QR-Code Detection Cerman Martin, 0625040 mcerman@prip.tuwien.ac.at
Content • Topic & Challenges (original proposal) • IDE‘s, Languages & Libraries • Chosen Approach • OutlierFiltering • Tuneable Parameters • TODO (original proposal / what was achieved) • Topic & Challenges (what was achieved) • Live Demonstration Android QR-Code Detection, Cerman Martin, 0625040
Topic & Challenges (original proposal) Topic • Detectandread QR-Codes on an Android Phone • As definedbythe ISO/IEC Standart • „Set ofblacksquares on a whitebackground“ • Varioussizes (Version 1-40) Challenges / Goals • Detectionundervariouslightingconditions • Perspectivedistortion • Detectionof QR-Code size • Real-Time Android QR-Code Detection, Cerman Martin, 0625040
IDE‘s, Languages & Libraries • MATLAB • Eclipse • ADT (Android Development Toolkit) • CDT (C++ Development Toolkit) • NDK (Android Native Development Kit) • Visual Studio 2012 • OpenCV Library Android QR-Code Detection, Cerman Martin, 0625040
Chosen Approach • „Fast Radial SymmetryforDetecting Points of Interest“, [1] • Algorithm • Transform tograyscaleandreduceimagesize • Determinegradient in x and y direction • Computeorientationandmagnitudeimage • Computesymmetryimageatcertainradii • Smooth using a Gausskernel • Non-maximumsuppression Android QR-Code Detection, Cerman Martin, 0625040
Chosen Approach (transformtograyscale) Android QR-Code Detection, Cerman Martin, 0625040
Chosen Approach (x derivative) Android QR-Code Detection, Cerman Martin, 0625040
Chosen Approach (y derivative) Android QR-Code Detection, Cerman Martin, 0625040
Chosen Approach (orientationimage) Android QR-Code Detection, Cerman Martin, 0625040
Chosen Approach (magnitudeimage) Android QR-Code Detection, Cerman Martin, 0625040
Chosen Approach (symmetryimage) Android QR-Code Detection, Cerman Martin, 0625040
Chosen Approach (thresholdedsymmetryimage) Android QR-Code Detection, Cerman Martin, 0625040
OutlierFiltering • Maximal numberofoutliersisaround 20 • Algorithm • Build a completegraph • Computeforeachvertexdistancesto all othervertices • Compute angle foreachvertextrio • Filter outliersbysettingrestrictions on maximal angle deviationanddistancedifferenceforeachvertextrio • Turned out toworkvery well Android QR-Code Detection, Cerman Martin, 0625040
Tuneable Parameters • Feature Detection • Image reductionfactor • Searchradius • Minimal numberofpositivelyaffectingpixels • Symmetrystrength • OutlierFiltering • Maximal edgelengthdifferenceofeachvertextrio • Maximal angle deviationofeachvertextrio • Frame Management • Split upsearching (different scales) andfilteringamongframes Android QR-Code Detection, Cerman Martin, 0625040
TODO (original proposal) • Determineparameters • Image size • Radii • Filter outliers • Based on neighborhood • Find suitablefeaturedescriptor • QR-Code reading Android QR-Code Detection, Cerman Martin, 0625040
TODO (what was achieved) • Determineparameters • Image size • Radii • Filter outliers • Based on neighborhood • Find suitablefeaturedescriptor • QR-Code reading Android QR-Code Detection, Cerman Martin, 0625040
Topic & Challenges (what was achieved) Topic • Detectandread QR-Codes on an Android Phone • As definedbythe ISO/IEC Standart • „Set ofblacksquares on a whitebackground“ • Varioussizes (Version 1-40) / Challenges / Goals • Detectionundervariouslightingconditions • Perspectivedistortion • Detectionof QR-Code size • Real-Time Android QR-Code Detection, Cerman Martin, 0625040
Live Demonstration your title
References • [1] Fast Radial SymmetryforDetecting Points of Interest, G. Loy andA. Zelinsky, IEEE Transactions on Pattern Analysis andMachineIntelligence, August 2003 Android QR-Code Detection, Cerman Martin, 0625040