Utilizing Super-Resolution for Enhanced Automotive Radar Object Detection
Asish Kumar Mishra, Kanishka Tyagi, Deepak Mishra
Abstract
In recent years, automotive radar has become an integral part of the advanced safety sensor stack. Although radar gives a significant advantage over a camera or Lidar, it suffers from poor angular resolution, unwanted noises and significant object smearing across the angular bins, making radar-based object detection challenging. We propose a novel radar-based object detection utilizing a deep learning-based super-resolution (DLSR) model. Due to the unavailability of low-high resolution radar data pair, we first simulate the data to train a DLSR model. We develop a framework that feeds a low-resolution radar dataset (called CRUW dataset) into the trained DLSR model pipeline to train a radar-based deep object detection classifier. The proposed framework achieves an 80% accuracy on object classification for the CRUW dataset and has a lower computational footprint, making it an ideal candidate for real-time implementation on edge devices used in autonomous driving applications. Code, dataset and supplementary material are on https://github.com/kanishkaisreal/DLSR_CRUW