Pavement Disease Image Classification Method

The automatic identification and classification of image-based pavement distress has been a challenge work in image processing and pattern recognition, and it plays a great role in the maintenance of highway. The background of my paper is a Jiangsu Province Natural Science Fund Project: research on road surface distress detection, classification and measurement..This paper focuses on the feature extraction methods and the classification methods. By analysis the texture of pavement distress binary images, we bring in the multi-resolution analysis method. We use the energy of seven sub-bands come from image's two-layer wavelet transform as the feature to describe the texture of the image. The support vector machine classifiers with different kernel functions are used to classify the feature. From the experiment, we can see that the feature extraction method is held to be valid. It can clearly distinguish five types of pavement distress and its accuracy is superior to Proximity and mixed density factor algorithms.As we see, the pavement distress detection and classification is a time consuming job, in order to meet the demands of practice, we provide a design of a simple and useful distributed computing platform which can engage several computers to run the distress detection system synchronously. The platform is based on a dynamic task allocation algorithm. Flexible architecture and load balance are the main characteristics of the distributed computing platform.