FecNet: A Feature Enhancement and Cascade Network for Object Detection Using Roadside LiDAR

Roadside light detection and ranging (LiDAR) is commonly used to record the traffic data of the whole intersection scene or road segment in intelligent transportation systems (ITSs). However, general deep-learning object detection methods do not adequately consider the static background captured by roadside LiDAR. Moreover, critical issues remain to be overcome in object detection using roadside LiDAR: false alarms caused by complex background interference and multiscale objects with limited characteristics. To this end, a feature enhancement and cascade network (FecNet) is proposed to alleviate the problems. From the perspective of feature enhancement, FecNet improves foreground feature discrimination by extracting foreground information and fusing it with feature maps of multiple stages. Also, from the perspective of feature cascade, a feature cascade backbone is proposed to enhance the localization and contextual information of multiscale objects with limited characteristics. Comprehensive experiments are conducted using a roadside LiDAR dataset. The experimental results suggest that FecNet is superior to the benchmark detectors and achieves better computational efficiency and detection accuracy.

I. Introduction

Road traffic safety is the key to realizing autonomous driving [1], [2], [3], [4], and sensing technology serves as a basis [5], [6]. However, the surrounding environment of autonomous vehicles is complex. For example, due to the limited perception range around the ego vehicle, onboard sensors have limited performance in preventing collisions at sharp bends or intersections. Different from onboard sensors, roadside sensors are installed at specific positions to obtain traffic data of the whole intersection or road segment. The fixed position enables them to avoid violent vibrations, thereby reducing data noise. Thus, using roadside sensors helps improve the safety of autonomous driving [7], [8].