Vi utbildar för framtiden och skapar samhällsnytta genom vår forskning som levandegörs i nära samarbete med näringslivet. Vi bedriver forskning inom computer science, datateknik, software engineering och interaktionsdesign - från grundforskning till direkta tillämpningar. Institutionen har en stark internationell prägel och är delad mellan Chalmers och Göteborgs universitet.
We are engaged in research and education across the full spectrum of computer science, computer engineering, software engineering, and interaction design, from foundations to applications. We educate for the future, conduct research with high international visibility, and create societal benefits through close cooperation with businesses and industry. The department is joint between Chalmers and the University of Gothenburg.
(2014) Martin, Johansson; Chalmers tekniska högskola / Institutionen för data- och informationsteknik (Chalmers); Chalmers University of Technology / Department of Computer Science and Engineering (Chalmers)
The purpose of this thesis has been to develop a automated pedestrian detection system for high mounted wide angle cameras on a Volvo truck, and prove its functionality for pedestrian detection in urban environments. Two cameras at different heights have been evaluated: one middle mounted forward facing camera; and one top mounted side facing camera. The thesis presents a framework of combining several independent algorithms fused into a unified decision for the presence of a pedestrian. For the forward facing camera a evaluation of image distortion was rst made. Two background subtraction techniques were implemented and compared: the Mixture of Gaussians, and the Codebook method. Further, two object classification methods were compared: the Viola & Jones method where salient features of pedestrians are captured by an overcomplete set of Haar-like wavelet features and chosen by the gentle AdaBoost training algorithm; and the continuation of this method by Dollar et al.. To handle the in-plane distortion a rotational scheme was evaluated and setup that resulted in five different classifier regions. The system was trained on pedestrian images captured around Lindholmen in Gothenburg, Sweden. Finally, tracking was incorporated in the form of Kalman filtering. For the side camera two new classifiers were trained based on the pedestrian bounding box viewing angle. An example interface was also developed that displays the final unified decision with a color bounding box warning system based on pedestrian proximity, the current active systems, a distance measure, and tracking history. Evaluation data was captured to test and verify the pedestrian detection and tracking method under normal urban environments. Experimental evaluation of the system on a conventional 2.16 GHz Intel Core2 CPU operating on 720*576 pixel images shows result of a robust detection and tracking system for pedestrians of different sizes, rotations and postures - with fast enough algorithms suitable for on-line operation.