Fraunhofer Center for Assistive Information and Communication Solutions – AICOSDescription:
The robustness of computer vision algorithms is strongly related with the quality of the images acquired. Besides the artificial or natural light exposure conditions, the consistency of handheld captures deeply depends on proper human interaction.
There are several solutions for light exposure and colour temperature control through enhanced camera parameterisation. However, these are not enough when a given use case requires the best criteria for the camera shoot decision based on the combination of correct geometry, accurate focus and fast segmentation during a mobile acquisition.
AICOS has been working on research and development of innovative solutions of practical utility in the areas of mHealth, mLogistics and mAgriculture. Some challenging use cases are currently requiring real-time feedback to the users, so that they can improve the handling of the smartphone during the acquisition, based on computationally low-cost approaches for quality control.
To tackle these challenges, real-time computer vision combined with inertial sensors on the smartphone, is expected to provide crucial data for the study of dedicated Human Computer Interaction during the capture.
In this project it is expected to explore new methods of visualization and interaction, using the interpreted data provided by AICOS research team gathered from camera and inertial sensors. The ultimate goal of this project is to study and implement a handheld image acquisition component for Android, which uses the real time image and signal features already developed, in order to achieve the best possible image acquisition.
Main Outcomes:
The main outcome will be the development of Handheld Image Acquisition component for Human-Computer Interaction (HCI) in Android smartphones. By using this component, AICOS research team will be able to:
> Build faster new android applicationss with handled acquisition system having a good level of abstraction;
> Identify and implement the best practises for handheld image acquisition and apply them in AICOS applications;
> Improve the performance of AICOS computer vision algorithms included in the applicationss;
> Solve some of the current problems related with mobile-based acquisition of insect traps (project DeMBACCHUS), mobile Optical character recognition (OCR) applications (project CounterFighting) and possibly, thanks to the new forms of user interaction, improve handheld acquisition of skin lesions (projects F3MImageAcquisitionModule, SkinsLesionsRS) or retinal lesions (project EyeFundusScope).
Author: Ricardo Meneses
Type: MSc thesis
Partner: Faculdade de Ciências da Universidade de Lisboa
Year: 2018
Download: