Disease Based On The Linear Form Of The Curvature Of The Road Image Information Detection Method

Pavement linear cracking detection is essential for the highway distress automatic detection, which is an important pre-treatment process. This paper studies the pavement cracking detection algorithms.In this paper, we study the detection algorithms from the aspects of differential geometry and make full use of the property of the curvature. We describe a new approach for detecting the pavement cracking, which extracts a wide range of linear distress information. The key point of our approach is to model the linear distress information as crest lines of the image surface. We achieve our target by extracting for the zero-crossings of the directional derivative of the maximum curvature which form crest lines. The image derivatives are computed using recursive filters approximating the Gaussian filter and its derivatives. The experiments show that the approach is effective, which is much higher than the classical edge detection in the detection accuracy. However, we consider the limitations of this approach. In the next paper, we show that, the extraction process of the linear distress information, having different width, requires a multi-scale analysis of the image. To design the fusion process of the multi-scale information, we will study the behavior of the curvatures in the scale space and the optimal scale. We illustrate the efficiency of the proposed multi-scale approach by extracting the linear distress information of different widths in the pavement images.Finally, we describe a clustering algorithm based on the curvature for the pavement images. Firstly we take full use of the detected image and extract the initial image information in these locations as a cluster of data sets. Then we choose the gray and the curvature as the factors of the K-means clustering. The experiments show that we weaken the false information and improve the image quality.