Identification of the part of soccer court from video signal by neural networks

Recently, it is becoming a popular practice to use character information in broadcasting video programs. This is useful for handicapped people with ears. In some cases, the characters are typed by typists. In that respect, it should be an important technology to produce the abstracted character information from the video information. The abstracted information will be useful for the normal audience, CG representation of the program, and store the game abstract in databases. In this paper, we will deal with the video processing part to identify the part of the court.