We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

eess

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Electrical Engineering and Systems Science > Signal Processing

Title: Beyond Point Clouds: A Knowledge-Aided High Resolution Imaging Radar Deep Detector for Autonomous Driving

Abstract: The potentials of automotive radar for autonomous driving have not been fully exploited. We present a multi-input multi-output (MIMO) radar transmit and receive signal processing chain, a knowledge-aided approach exploiting the radar domain knowledge and signal structure, to generate high resolution radar range-azimuth spectra for object detection and classification using deep neural networks. To achieve waveform orthogonality among a large number of transmit antennas cascaded by four automotive radar transceivers, we propose a staggered time division multiplexing (TDM) scheme and velocity unfolding algorithm using both Chinese remainder theorem and overlapped array. Field experiments with multi-modal sensors were conducted at The University of Alabama. High resolution radar spectra were obtained and labeled using the camera and LiDAR recordings. Initial experiments show promising performance of object detection using an image-oriented deep neural network with an average precision of 96.1% at an intersection of union (IoU) of typically 0.5 on 2,000 radar frames.
Comments: 6 pages, 9 figures
Subjects: Signal Processing (eess.SP)
Cite as: arXiv:2111.01246 [eess.SP]
  (or arXiv:2111.01246v1 [eess.SP] for this version)

Submission history

From: Shunqiao Sun [view email]
[v1] Mon, 1 Nov 2021 20:17:52 GMT (9347kb,D)

Link back to: arXiv, form interface, contact.