Litcius/Paper detail

Unstructured Road Vanishing Point Detection Using Convolutional Neural Networks and Heatmap Regression

Yinbo Liu, Ming Zeng, Qing‐Hao Meng

2020IEEE Transactions on Instrumentation and Measurement22 citationsDOI

Abstract

Unstructured road vanishing point (VP) detection is a challenging problem, especially in the field of autonomous driving. In this article, we proposed a novel solution combining the convolutional neural network (CNN) and heatmap regression to detect unstructured road VP. The proposed algorithm first adopted a lightweight backbone, i.e., depthwise convolution-modified HRNet, to extract hierarchical features of the unstructured road image. Then, three advanced strategies, i.e., multiscale supervised learning, heatmap superresolution, and coordinate regression techniques were utilized to carry out fast and high-precision unstructured road VP detection. The empirical results on Kong's data set showed that our proposed approach had the highest detection accuracy in real-time compared with the state-of-the-art methods under various conditions, and achieved the highest speed of 33 fps.

Topics & Concepts

Convolutional neural networkComputer scienceConvolution (computer science)Artificial intelligencePattern recognition (psychology)RegressionSet (abstract data type)Image (mathematics)Point (geometry)Artificial neural networkMathematicsStatisticsProgramming languageGeometryAutonomous Vehicle Technology and SafetyInfrastructure Maintenance and MonitoringAdvanced Neural Network Applications
Unstructured Road Vanishing Point Detection Using Convolutional Neural Networks and Heatmap Regression | Litcius