TY - GEN
T1 - Real-time electric vehicle classification for electric charging and parking system using pre-trained convolutional neural network
AU - Seo, Yian
AU - Shin, Kyung shik
N1 - Publisher Copyright:
© 2019 Association for Computing Machinery.
PY - 2019
Y1 - 2019
N2 - Deep learning can be applied on vehicle image classification diverse purposes such as real-time vehicle recognition, vehicle license plate character recognition, vehicle logo identification, and vehicle color classification. We categorize electric vehicle model images according to electric vehicle types and classify each image using Convolutional Neural Network for electric vehicle parking and charging system. The suggested model can support optimization problem of charging equipment installation and usage by recognizing the electric vehicle type real-time basis at the entrance of parking or charging system and directing each vehicle where to park and charge as each has different charge method and needs different charging equipment. The model is built using transfer learning, that is, we pre-train the network with large dataset compensating the small amount of main dataset and then fine-tune the network with our electric vehicle type dataset. We perform 10-fold experiments and achieve final test accuracy of 77.6%.
AB - Deep learning can be applied on vehicle image classification diverse purposes such as real-time vehicle recognition, vehicle license plate character recognition, vehicle logo identification, and vehicle color classification. We categorize electric vehicle model images according to electric vehicle types and classify each image using Convolutional Neural Network for electric vehicle parking and charging system. The suggested model can support optimization problem of charging equipment installation and usage by recognizing the electric vehicle type real-time basis at the entrance of parking or charging system and directing each vehicle where to park and charge as each has different charge method and needs different charging equipment. The model is built using transfer learning, that is, we pre-train the network with large dataset compensating the small amount of main dataset and then fine-tune the network with our electric vehicle type dataset. We perform 10-fold experiments and achieve final test accuracy of 77.6%.
KW - Convolutional Neural Network
KW - Electric charging
KW - Electric vehicle
KW - Transfer learning
KW - Vehicle classification
UR - http://www.scopus.com/inward/record.url?scp=85066945156&partnerID=8YFLogxK
U2 - 10.1145/3322645.3322650
DO - 10.1145/3322645.3322650
M3 - Conference contribution
AN - SCOPUS:85066945156
SN - 9781450361033
T3 - ACM International Conference Proceeding Series
SP - 8
EP - 11
BT - ACM International Conference Proceeding Series
PB - Association for Computing Machinery
Y2 - 16 March 2019 through 19 March 2019
ER -