Software Developer

• Porting and extending an algorithm for detecting a covered surveillance camera to a current SDK
• Implementation of new features for a video management system

• Implementation of filters for control software that operates laser welding systems
• Analysis of an existing algorithm for creating a large photograph from multiple smaller images
• Development of measures to optimize the algorithm and minimize artifacts

• Implementation of a QA software for optical monitoring of the powder bed of a 3D metal printer using a camera
• Performing image enhancement measures to document process steps

• Implemented a prototype to read an ID card using a camera and prepare it for OCR (Tesseract)
• Developed a tool to adjust the algorithm for different ID cards

• Ported monitoring algorithms to Linux embedded cameras (HikVision, Hanwha, Axis)
• Configured Eclipse for different targets
• Created shell scripts to automate build processes
• Integrated TFS into the Linux development environment
• Designed a unified architecture for various camera manufacturers
• Built a cross-platform library for ARM, Intel, Linux, and Windows
• Established an efficient desktop development and debugging environment
• Streamed video from the camera into the desktop development environment

• Checking the print quality of laser-etched photos including gray value calibration, localization via correlator, precise scaling, and color space transformation (LBP, Canny, Sobel, HSV)
• Comparing reference and etched photos and classifying detected defects

Training:

• Generation of artificial characters in various fonts
• Scaling artificial characters to a uniform size
• Calculation of gradients and training of a support vector machine

Recognition:

• Searching for blobs in the image
• Scaling the found blobs
• Calculating the gradients and classification with the support vector machine

• Ensuring print quality of the Machine Readable Zone (MRZ) in passports according to ISO 1831 using OCR-B TrueType font templates
• Checking contrast, stroke width, smudges, alignment and size of the characters

Training:

• Segmentation of perforated characters (dots) using adaptive thresholding
• Normalization and scale calculation
• Capturing coordinates per character (5–10 per character)

Detection:

• Segmentation and clustering of dots using dilate + erode
• Translation and normalization of the dots
• Rotation estimation for matching with the training phase
• Algorithm is scale- and rotation-invariant, has no error detection and is flexibly configurable

Algorithm 1:

• Created a 3D model and on-the-fly key-point calculation including 3D coordinates
• Runtime key-point search and matching via RANSAC for position estimation

Algorithm 2:

• Created a geometric line model and line detection using Hough transform
• Search for a matching transformation and verify line fitting

• Integration and control of AVT industrial cameras in a software project

• Acceleration of markerless camera calibration with Bundler by parallelizing and splitting videos
• Splitting videos into overlapping parts and computing external camera parameters and transforming them into a common coordinate system

• Implementation of the DAB-MOT protocol and a GUI for visualizing stock market data in Java
• Retrieval of real-time stock market data via web service and feeding it into the DAB signal via NDR