master thesis

SignatureMA 018
AuthorHang Yu
TitleÜberwachung des Gefahrenraumes zwischen Bahnsteigtür und Zug in einem Metro-System mittels digitaler Bildverarbeitung
TutorDipl.-Ing. Bertolt Blümke, Dipl.-Ing. Volker Eiselein
ProfessorDr.-Ing. Thomas Sikora
AbstractBackground subtraction means that the dynamic or static objects are segmented from image sequence and the lightweight background is eliminated. In recent years it is referred as one of the key technologies in the field of computer vision and object tracking. This thesis proposes a particularly effective background subtraction algorithm combined with a linear classifier for the detection of unwanted objects in the danger zone between platform door and train door. This background subtraction algorithm is based on HSV colour space, applies block processing and linear dependence of vectors in terms of value. Comparing with other background subtraction algorithms, which have been introduced in this thesis, this algorithm shows the superiority in terms of accuracy, speed and stability during brightness changing and shadows occurring. Furthermore, a linear classifier is designed with the feature - the area of the segmented foreground - through studying with the "minimum mean square error" method. The test result proves that the whole algorithm implement the detection of objects in the danger zone between platform door and train door successfully within the required time period.
Key wordsbackground subtraction, HSV colour space, block processing, linear dependence of vectors, linear classifier, minimum mean square error method