Computer vision and machine learning for the detection and classification of pavement cracks.

For road maintenance and rehabilitation, it is important to develop procedures to evaluate pavement condition. Imaging methods can be used to obtain data to analyze a pavement surface. A methodology for crack detection is presented in this paper that is based on image processing techniques and artificial neural networks. The methodology is implemented in four stages: 1. image acquisition, 2. image processing, 3. feature extraction, and 4. classification using an artificial neural network. The methodology was used to detect deterioration in the form of longitudinal cracks, potholes, and alligator cracking. The classification was performed using a multilayer perceptron (MLP) neural network within a (12 14 3) configuration, resulting in an accuracy of 95.56% and a precision of 94.44%. The proposed methodology could be used to help governmental organizations evaluate a road network. [ABSTRACT FROM AUTHOR]