Abstract

The investigation of tire-road noise according to the type of road pavement is time-consuming and expensive. In this study, an artificial neural network model was applied to address this problem. Models to classify road pavement types (for example, transverse-tined, longitudinal-tined, NGCS, DG, and SMA) were implemented and their performance were compared. Input data were constructed by combining the features extracted from tire-road noise and road surface images. The tire-road noise collected using the OBSI measurement method was analyzed for the sound pressure level, sound intensity, and sound quality indices. Road surface image data were analyzed using the image feature extraction algorithms of the Hough transformation and histogram of gradient(HOG). The top 10 important variables were selected by inputting each feature into a random forest model, and artificial neural network models were constructed by each feature. The classification accuracy of the model using only acoustic features was 90.8 % and that using only image features was 88.8 %. The accuracy of the model using both features was 97.3 %. The overall classification performance was improved by using the acoustic and image features.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call