Road roughness recognition based on improved fuzzy C-mean algorithm combined with genetic algorithm

Abstract: With the development of Chinese economy and the increasing instruction of highway, road detection and identification has become the focus for infrastructure, thus the requirements of the road roughness detection accuracy are also urgent. In road recognition cases, due to different road load spectrum has obvious clustering features, it is feasible for road recognition by using clustering analysis method. However, clustering result is sensitive to the initial center, and often cannot achieve the result of the global optimal. FCM (fuzzy C-mean) clustering algorithm and improved FCM algorithm were used to measure the effect of clustering by using deterministic objective function, the existence of the local minimum points of the objective function made clustering result sensitive to the initial center. So FCM algorithm and improved FCM algorithm both have shortcomings and cannot solve the problem that the effect of clustering is bad when sample contains noise. In this paper a new method of improved FCM clustering algorithm combined with genetic algorithm was proposed. The method mainly included the following four steps: Firstly, genetic algorithm was adopted to get the clustering center. Secondly, improved FCM clustering algorithm was used to get the optimal solution. Thirdly, the prominent IRIS data was used to validate its effectiveness. Lastly, improved FCM algorithm combined with genetic algorithm was applied to the real road surface spectrum data for the identification of pavement in the project. IRIS data were used to verify the clustering performance of fuzzy kernel clustering algorithm in this paper. It confirmed that improved FCM algorithm effectively solved the problem of the bad effects of clustering when the sample contained much noise and improved the effectiveness of clustering. In proposed method, a non Euclidean distance was used to replace the Euclidean distance of FCM algorithm in order to reduce the influence of noise, and improved FCM algorithm was combined with genetic algorithm so as to avoid the problem that FCM algorithm and its improved algorithm were sensitive to initial value and easy to fall into local optimal solution. The method had the following advantages: on the one hand it can distinguish, extract and amplify useful features effectively; on the other hand it was clustered in the feature space so as to get a better classification result. Experimental results showed that the membership degree of improved FCM algorithm was more reasonable than original FCM algorithm, and it also showed that the improved FCM algorithm combined with genetic algorithm had a better ability to deal with noise data and was much better performance than improved FCM algorithm and original FCM algorithm, the accuracy of classification was improved. The road recognition rate of the improved clustering algorithm was up to 94.54%, which was 4.98 percent points higher than FCM algorithm, 4.67 percent points higher than improved FCM algorithm. It can be seen from the above data that the new clustering algorithm greatly improves the accuracy of clustering and the road recognition rate. The new algorithm can be effectively applied to road classification.