Effects of the Feature Extraction from Road Surface Image for Road Induced Noise Prediction Using Artificial Intelligence Toyota Motor Corporation

Format:

Book

Conference/Event

Author/Creator:

Nakamura, Nakamura, author.

Contributor:

Ishizaki, Keisuke

Komada, Masashi

Matsumura, Yuichi

Matsushita, Kojiro

Conference Name:

Noise and Vibration Conference & Exhibition (2019-06-10 : Grand Rapids, Michigan, United States)

Language:

English

Physical Description:

1 online resource cm

Place of Publication:

Warrendale, PA SAE International 2019

Summary:

AbstractNext generation vehicles driven by motor such as electric vehicles and fuel cell vehicles have no engine noise. Therefore the balance of interior noise is different from the vehicles driven by conventional combustion engine. In particular, road induced noise tends to be conspicuous in the low to middle vehicle speed range, therefore, technological development to reduce it is important task. The purpose of this research is to predict the road induced noise from the signals of sensors adopted for automatic driving for utilizing the prediction result as a reference signal to reduce road induced noise by active noise control (ANC). Using the monocular camera which is one of the simplest image sensors, the road induced noise is predicted from the road surface image ahead of the vehicle by machine learning. The effects to extract features (Histograms of Oriented Gradients (HOG) feature, autoencoder feature, Convolutional Neural Network (CNN) feature) from road surface images are evaluated by visualization result of t-SNE. From the features acquired by the above method, the frequency characteristics of the road induced noise are predicted using deep learning, which is generally considered to be high prediction accuracy as a machine learning method. For the eight kinds of road surface, we compared the road induced noise measured by the actual vehicle and predicted by using deep learning. The prediction result using extracted features of road surface image is more accurate than that using RGB value of road surface directly. This tendency accorded with a visualization result by t-distributed Stochastic Neighbor Embedding (t-SNE)

Notes:

Vendor supplied data

Publisher Number:

2019-01-1565

Access Restriction:

Restricted for use by site license