The invention discloses a rapid arc detection method. The method is characterized by comprising steps of 1, edge image acquisition; 2, decomposition of edges into multiple candidate arcs; 3, arc combination; and 4, arc screening and detection. According to the invention, the method is fast in speed of the adopted detection algorithm, highly in precision, high in robustness and widely applicable to detection of all kinds of objects including arc boundaries in the field of machine vision.