Lane Polynomial Regression with Dual-Dimension Attention Convolution

Conference: ICMLCA 2021 - 2nd International Conference on Machine Learning and Computer Application
12/17/2021 - 12/19/2021 at Shenyang, China

Proceedings: ICMLCA 2021

Pages: 7Language: englishTyp: PDF

Personal VDE Members are entitled to a 10% discount on this title

Authors:
Chen, Lulu; Weng, Xiaoxiong (School of Civil Engineering and Transportation, South China University of Technology, Guangzhou, China)
Luo, Ruifa (Shenzhen Genvict Technologies Co., Ltd. Shenzhen, China)
Cao, Yu (College of Information Science and Engineering, China University of Petroleum, Beijing, China)
Yilihamu, Abudujiasuer (Universite de Tours Polytech, France, Tours, France)

Abstract:
Advanced driving assistance systems such as lane departure warning and adaptive cruise control have been applied to smart vehicles. Lane detection is the basis of these applications. Accurate and fast lane detection can realize vehicle positioning in time and assist in making appropriate decisions. Vision-based lane detection has gradually become the mainstream method because of its accuracy and higher cost performance. However, the lanes are slender objects in a picture taken from the camera in front of a vehicle. Therefore, it is important to grasp the weight of each position and the response of the lane category in lane detection. This paper goes beyond the aforementioned limitations and proposed a lane regression network with dual-dimensional attention convolution. On the one hand, it prompted CNN to seize the global information that the corresponding spatial and channel attention map is generated to assign different weights to the input feature. On the other hand, the lane polynomial coefficients regression ensures the speed of lane detection. Finally, the model proposed in this paper was verified on the TuSimple lane detection benchmark and achieved competitive results. At the same time, convolutional visualization technology was used to verify the effectiveness of the attention mechanism.