A new method to pavement cracking detection based on the Biological Inspired Model

Abstract

Due to the complexity of shape and apparent differences of pavement cracks, it is difficult to characterize them with definite features. The wavelet, Gabor transform and its functions are usually predefined and cannot adapt to the characteristics of the pavement crack images. This paper proposes a novel joint maximization recognition algorithm in the resilient area, which is based on the characteristics of biologically inspired model (BIM). In view of the predefined and invariance of the basis functions in linear transformation, this algorithm uses the resilient area, such as the four regions or eight areas to segmenting image. Introducing Adaboost classifier in each area to select and retain key information, get rid of unwanted or negative information. Its eigenvectors can reflect the information in the original image comprehensively and its low computational complexity and parallelizable is helpful in real-time applications. The experimental results show that the overall recognition rate of the proposed method in pavement cracks is up to 99.23%, and its fast response time fully demonstrate the effectiveness of this method.