Machine Learning Based Early Debris Detection Using Automotive Low Level Radar Data
Kanishka Tyagi, Shan Zhang, Yihang Zhang, John M. Kirkwood, Sanling Song, Narbik Manukian
Abstract
Road safety for automated vehicles requires accurate and early detection of stationary objects in the vehicle’s path. Radar can use doppler to effectively identify stationary objects and make these identifications at long range and in severe weather and poor light conditions. In this paper, we propose a radar-based stationary object detection system that combines signal processing techniques with machine learning technology to detect stationary in-path objects from the low level spectra of front looking radars. The proposed system consists of novel signal and image processing methods to extract key features from the raw data, which are fed into a long short-term memory (LSTM) to determine the probability of a stationary object in-lane at each range. Experiments with collected data in controlled and uncontrolled scenarios demonstrate the effectiveness of our approach.