Job Title
Master Thesis – Pan Tilt Zoom Cameras: Two Different Ideas
Job Description
Image Processing & Zoom Drift compensation for PTZ Cameras (Choose between two tracks)
Category
Computer Science, Image Processing, Embedded Systems, Machine Learning, Optics, Mechatronics
Scope
Master Thesis, 30 hp (ECTS). The project is intended for two students working together.
Below, we present two potential project tracks within our PTZ Imaging team. As applicants, you can indicate in your application which topic interests you the most, or if you are open to either. The final scope will be chosen together with the team before starting.
Track 1: Real-time Blemish Detection and Mitigation in Sky-tracking Applications
Background
Axis PTZ (pan/tilt/zoom) cameras are increasingly used in demanding scenarios, such as tracking objects against the sky. In these cases, minor imperfections on the sensor or lens—often called "blemishes"—can appear under specific lighting conditions. These blemishes risk being misinterpreted as moving targets by analytics algorithms, causing false positives. To ensure reliable performance, we need a method to identify and handle these blemishes directly on the camera in real time.
Goal & Key Parts
The goal is to investigate, implement, and evaluate methods for real-time blemish detection in a video stream from a PTZ camera.
Literature study: Survey existing methods for detecting image sensor defects and optical artifacts.
Algorithm development: Design a robust detection algorithm capable of handling PTZ movements and changing sky lighting.
Proof-of-Concept: Implement the algorithm as a functional prototype, ideally running directly on an Axis camera platform.
Evaluation: Benchmark detection accuracy and computational efficiency.
Track 2: Real-time Compensation for Optical Center Shift in Long-Zoom PTZ Cameras
Background
Axis long-zoom PTZ cameras provide powerful surveillance capabilities across vast distances. However, at long zoom levels, mechanical tolerances in lens assembly and calibration can cause the optical center to drift across the sensor in the xy-plane during zoom. This creates issues for coordinate-dependent features such as privacy masks, which risk shifting and exposing sensitive areas. A real-time method is needed to counteract this drift.
Goal & Key Parts
The goal is to develop and evaluate a method to actively compensate for optical center shift during zoom.
Characterization & Mapping: Develop a methodology to map xy-drift as a function of the zoom level for a specific camera unit.
Real-time Compensation: Design and implement a method to compensate for the drift.
Implementation: Deploy the compensation mechanism as a proof-of-concept on an Axis camera platform.
Evaluation: Measure compensation precision, computational overhead, and stability of privacy masks.
Who are you?
We are looking for two students enrolled in a Master’s program in Computer Science, Engineering Physics, Electrical Engineering, Mechatronics, or a related field. You have a keen interest in image processing and the intersection of hardware and software.
Requirements:
Strong programming skills in C++ and/or Python.
A solid foundation in image processing and coordinate transformations/analysis.
Appreciated Qualifications:
Familiarity with embedded systems, edge devices, or real-time programming.
Experience with machine learning for computer vision (Track 1).
Knowledge of camera optics, lens mechanics, or control theory (Track 2).
Practical experience with hands-on hardware-software integration.
OK, I am interested! What do I do now?
You are valuable to us – how nice that you are interested in our proposal! There are a few things to keep in mind when applying:
Applications are accepted in both Swedish and English via the proposal advert.
The announced thesis is open only to students affiliated with a Swedish University/College either directly or via an exchange program.
Since the project involves two students working together,…