ICanC: Improving Camera-based Object Detection and Energy Consumption in Low-Illumination Environments

• Daniel Ma, Computer Science, Northeastern University

• Weisong Shi, Computer and Information Sciences, University of Delaware
• Ren Zhong, Computer Science, Wayne State University

This paper introduces ICanC, a novel system de-

signed to improve object detection and reduce energy consumption

in autonomous vehicles (AVs) operating in low-illumination

environments. By leveraging the complementary strengths of

LiDAR and camera sensors, ICanC enhances detection accuracy

in conditions where camera performance typically falters while

minimizing unnecessary headlight usage, thus supporting the

overarching goal of sustainable transportation.

ICanC comprises three main nodes: the Obstacle Detector,

which fits bounding boxes onto objects detected in the LiDAR

sensor’s point cloud and calculates pertinent information such

as position, velocity, and orientation; the Danger Detector, which

assesses potential threats from these bounding boxes; and the

Light Controller, which activates headlights to improve camera

visibility only when danger is detected.

Data collected from both physical and simulated environments

demonstrate that ICanC performs as intended, even in the

abundant presence of noise. Although the system imposes stress

on computing resources, it maintains high accuracy in camera-

based object detection assisted by headlights, while potentially

significantly reducing headlight energy usage. ICanC offers a

promising approach to AV research, improving upon energy

conservation while simultaneously maintaining accurate detection.