Vehicle Type Recognition Based on Radon-CDT Hybrid Transfer Learning
Shujuan Guan, Bin Yih Liao, You Ru Du, Xiangyun Yin · 2019
Recognition of vehicle type is an important and challenging tasks. To improve the accuracy and reduce the computational complexity of model for vehicle type recognition, this paper proposes the vehicle type recognition method based on Radon-CDT hybrid transfer learning. Deep features of images are obtained by the pre-trained CNNs, then Radon-CDT are used to capture the non-linearly property of the features. This method makes full use of the excellent feature extraction ability of convolution layer and the property to make data linearly separable of Radon-CDT. Experimental results show that the proposed method is a very promising alternative to deal with the recognition of vehicle type, which obtains a better recognition accuracy than other popular models (AlexNet, VGG and ResNet).