Pavement damage type identification method based on classifier ensemble

Abstract

The present invention discloses a pavement damage type identification method based on a classifier ensemble. The method comprises the following steps of: 1) performing homomorphic filter preprocessing on detected damaged pavement images; 2) establishing a pavement damage image database by using the preprocessed pavement images, and guaranteeing the balance of image samples in the database, namely, guaranteeing the images to be same in size and number, wherein the database comprises horizontal crack, longitudinal crack, net crack and precipitation damage types; 3) extracting eigenvectors of the pavement damage images by respectively using Contourlet conversion and edge direction histogram methods, and establishing union eigenvectors of the pavement damage images by using a serial combined policy; and 4) based on a cross kernel support vector machine, establishing a random subspace ensemble classifier to identify the pavement damage images. The pavement damage type identification method based on the classifier ensemble, provided by the present invention, effectively improves the efficiency of pavement damage image type identification,and provides an efficient and intelligent management basis for road management work, particularly pavement maintenance.

Description

technical field [0001] The patent of the present invention relates to the field of intelligent transportation research, especially the automatic identification method of pavement damage type in pavement maintenance and management. Background technique [0002] The increase in traffic volume, large-scale vehicles, serious overloading and other phenomena lead to the appearance of road surface damage, which seriously affects the safety and comfort of road traffic. Pavement maintenance is the focus of highway maintenance, and pavement damage status plays an important role in pavement maintenance decision-making as the basis of pavement maintenance management. The types of pavement images include transverse cracks, longitudinal cracks, network cracks and normal types. The traditional pavement damage detection method uses manual detection, but this method has low work efficiency, high labor intensity, high degree of operation risk, and is affected by weather conditions, etc. The ...

AI Technical Summary

Problems solved by technology

In order to effectively realize the intelligent recognition of pavement damage, many research institutions and scholars have carried out related research work. At present, the feature description of pavement damage images is mostly used to describe the area and geometric features. Sun Ao et al. A method for classifying pavement damage by using support vector machine, but this method is only suitable for crack-type damage, and the classification effect for non-crack-type irregular damage is weak

Images