Detecting crack type and crack size is crucial for road maintenance and management. Mobile crowd sensing is a new way to collect the information of cracks on roads. We propose a system named CrackDetector to detect cracks and estimate their types and size with smart phone in this paper. The type of a crack (i.e., horizontal crack, vertical crack, net crack) is determined by a coordinate transmission method based on three directions, including the direction of the crack in the photo (estimated through an image processing method), the 3D shooting direction (i.e., the rotation of the smartphone while photographing) and the direction of the road (obtained from the OpenStreetMap road network). In order to estimate the size of a crack (i.e., the width and the length of the crack), we use the camera’s convex lens imaging theory and readings of the accelerometer and the magnetometer sensor. We collected pictures of 152 cracks with a specifically-developed android application by 8 volunteers. Experimental results show that our approach achieves an accuracy of 90.1% in crack type detection. Meanwhile, we get a 3.2cm RMSE error when estimating the crack width and a 13.2cm RMSE error when estimating the crack length.