ColorCrack: Identifying Cracks in Glass

Automatically identifying cracks in glass is an application of machine vision that has not been studied much before. It poses the unique opportunity to incorporate the physics of glass and how it cracks into a detection algorithm. ColorCrack is an automated method of finding cracks in glass smartphone screens using insights from previous work in object and edge detection.The key to ColorCrack is the ability to automatically detect cracked screens from nearly any angle. This functionality allows the system to take advantage of the fact that cracks are more visible at certain angles to improve the detection accuracy. The algorithm was developed and evaluated using a personally collected dataset. Initial results show that ColorCrack can typically find the screen correctly or with little error, as well as identify the main crack structures, efficiently enough for most applications. In the future, the ColorCrack algorithm could be extended to other glass types for applications and has uses in fields such as material and forensic science.