Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

High Dynamic Range Modulo Imaging for Robust Object Detection in Autonomous Driving

Kebin Contreras, Brayan Monroy, Jorge Bacca

TL;DR

This paper tackles saturation-driven information loss in autonomous driving object detection by introducing modulo sensors that encode irradiance with wrap-around, enabling HDR-like detail without multi-exposure capture. It couples modulo imaging with the SPUD HDR recovery algorithm to obtain $\hat{x}$ and evaluates detection performance with YOLOv10 on KITTI, showing results comparable to ideal HDR and superior to saturated images while reducing HDR acquisition time. The two-step pipeline—modulo acquisition with SPUD-based HDR recovery followed by detection on either $y$ or $\hat{x}$—achieves real-time robustness under extreme lighting and without retraining. Overall, the approach offers a practical route to reliable, fast object detection in challenging illumination for autonomous driving applications.

Abstract

Object detection precision is crucial for ensuring the safety and efficacy of autonomous driving systems. The quality of acquired images directly influences the ability of autonomous driving systems to correctly recognize and respond to other vehicles, pedestrians, and obstacles in real-time. However, real environments present extreme variations in lighting, causing saturation problems and resulting in the loss of crucial details for detection. Traditionally, High Dynamic Range (HDR) images have been preferred for their ability to capture a broad spectrum of light intensities, but the need for multiple captures to construct HDR images is inefficient for real-time applications in autonomous vehicles. To address these issues, this work introduces the use of modulo sensors for robust object detection. The modulo sensor allows pixels to `reset/wrap' upon reaching saturation level by acquiring an irradiance encoding image which can then be recovered using unwrapping algorithms. The applied reconstruction techniques enable HDR recovery of color intensity and image details, ensuring better visual quality even under extreme lighting conditions at the cost of extra time. Experiments with the YOLOv10 model demonstrate that images processed using modulo images achieve performance comparable to HDR images and significantly surpass saturated images in terms of object detection accuracy. Moreover, the proposed modulo imaging step combined with HDR image reconstruction is shorter than the time required for conventional HDR image acquisition.

High Dynamic Range Modulo Imaging for Robust Object Detection in Autonomous Driving

TL;DR

This paper tackles saturation-driven information loss in autonomous driving object detection by introducing modulo sensors that encode irradiance with wrap-around, enabling HDR-like detail without multi-exposure capture. It couples modulo imaging with the SPUD HDR recovery algorithm to obtain and evaluates detection performance with YOLOv10 on KITTI, showing results comparable to ideal HDR and superior to saturated images while reducing HDR acquisition time. The two-step pipeline—modulo acquisition with SPUD-based HDR recovery followed by detection on either or —achieves real-time robustness under extreme lighting and without retraining. Overall, the approach offers a practical route to reliable, fast object detection in challenging illumination for autonomous driving applications.

Abstract

Object detection precision is crucial for ensuring the safety and efficacy of autonomous driving systems. The quality of acquired images directly influences the ability of autonomous driving systems to correctly recognize and respond to other vehicles, pedestrians, and obstacles in real-time. However, real environments present extreme variations in lighting, causing saturation problems and resulting in the loss of crucial details for detection. Traditionally, High Dynamic Range (HDR) images have been preferred for their ability to capture a broad spectrum of light intensities, but the need for multiple captures to construct HDR images is inefficient for real-time applications in autonomous vehicles. To address these issues, this work introduces the use of modulo sensors for robust object detection. The modulo sensor allows pixels to `reset/wrap' upon reaching saturation level by acquiring an irradiance encoding image which can then be recovered using unwrapping algorithms. The applied reconstruction techniques enable HDR recovery of color intensity and image details, ensuring better visual quality even under extreme lighting conditions at the cost of extra time. Experiments with the YOLOv10 model demonstrate that images processed using modulo images achieve performance comparable to HDR images and significantly surpass saturated images in terms of object detection accuracy. Moreover, the proposed modulo imaging step combined with HDR image reconstruction is shorter than the time required for conventional HDR image acquisition.

Paper Structure

This paper contains 5 sections, 5 equations, 3 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Background
  3. Methodology
  4. Simulation and Results
  5. Conclusion

Figures (3)

  • Figure 1: Challenges of sensor saturation in autonomous driving environments from various databases: (a) BDD100K shows problems of overexposure due to direct sunlight yu2020bdd100k; (b) Nighttime Driving depicts reflections from artificial night lights daytime:2:nighttime; and (c) KITTI demonstrates the loss of road details due to solar saturation Geiger2012CVPR.
  • Figure 1: Comparison of different imaging techniques under varying levels of light saturation. The columns represent three different methods: Saturated, Modulo, and Recovery. The rows illustrate the results for different values of the scaling factor $\alpha$, ranging from 1.5 to 8.
  • Figure 2: Comparison of imaging techniques under varying levels of light saturation. Each column stands for; Ideal HDR, Saturated, Modulo, and Recovery. Each row shows the results for different values of the saturation factor $\alpha$, ranging from 1.5 to 4.