Computer Vision Based Crack Detection and Analysis

Cracks on a bridge deck should be ideally detected at an early stage in order to prevent further damage. To ensure safety, it is necessary to inspect the quality of concrete decks at regular intervals. Conventional methods usually include manual inspection of concrete surfaces to determine defects. Though very effective, these methods are time-inefficient. This paper presents the use of computer-vision techniques in detection and analysis of cracks on a bridge deck. High quality images of concrete surfaces are captured and subsequently analyzed to build an automated crack classification system. After feature extraction using the training set images, statistical inference algorithms are employed to identify cracks. The results demonstrate the feasibility of the proposed crack observation and classification system.