| Resumo: | Object detection in digital images with small computational costs and processing time is a necessity of some of the most diverse domains, such as: robotics, video surveillance or traffic analysis, among others. With current advances in technology and the decrease of prices in image sensors and video cameras, nowadays digital cameras can acquire images at high resolutions and frame rates. This means that a great amount of visual information has to be processed in a small amount of time. This Ph.D. addresses the problem of time-constrained detection of colored objects and uses as a test bed intelligent robots that play soccer in an autonomous manner. This work is integrated in the Intelligent Robotics and Systems Laboratory from the Institute of Electronics and Informatics Engineering of Aveiro of University of Aveiro. The autonomous robotic soccer team CAMBADA was the test bed of all the work developed during this Ph.D. The objective was to present a modular solution of a vision system that could be used for time-constrained color detection in the aforementioned robotic project. Algorithms for colormetric camera calibration, optimal image scanning and colored blob formation have been designed and based on them functional robotic vision systems have been developed. These vision systems have been used in real adversarial soccer scenarios and the results obtained prove their efficiency. Moreover, all the developed algorithms and software have been aggregated within a computer vision library, which is both free and open source and designed for time-constrained colored object detection. In addition, several support tools have been implemented as well, as tools for remote monitoring or color calibration of a robotic vision system.
|
|---|