Long exposure localization in darkness using consumer cameras

2013-05-06

IEEE International Conference on Robotics and Automation

AI Summary

This work addresses the challenge of reliable visual localization using low-cost cameras under extremely low-light conditions—two orders of magnitude darker than standard benchmarks—where prolonged exposure and high ISO introduce severe motion blur, degrading conventional methods. We systematically evaluate SeqSLAM’s robustness under such extreme blur. Methodologically, we acquire usable grayscale images via long-exposure (132–10,000 ms) and high-gain imaging, and enhance SeqSLAM’s blur tolerance through block-wise and local-neighborhood normalization. We provide the first mechanistic insight into SeqSLAM’s effectiveness under strong motion blur and empirically validate its cross-illumination and cross-perceptual-domain generalization—e.g., daytime training to nighttime localization. Experiments demonstrate stable localization in both synthetic and real-world ultra-low-light scenarios. Statistical analysis confirms that normalization is critical for maintaining robustness against motion blur.

Technology Category

• Computer Vision: Low Level & Physics-based Vision
• Intelligent Robots: Localization, Mapping, and Navigation
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Security and Privacy: Large-scale security measurements
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

In this paper we demonstrate passive vision-based localization in environments more than two orders of magnitude darker than the current benchmark using a $100 webcam and a $500 camera. Our approach uses the camera's maximum exposure duration and sensor gain to achieve appropriately exposed images even in unlit night-time environments, albeit with extreme levels of motion blur. Using the SeqSLAM algorithm, we first evaluate the effect of variable motion blur caused by simulated exposures of 132 ms to 10000 ms duration on localization performance. We then use actual long exposure camera datasets to demonstrate day-night localization in two different environments. Finally we perform a statistical analysis that compares the baseline performance of matching unprocessed grayscale images to using patch normalization and local neighborhood normalization - the two key SeqSLAM components. Our results and analysis show for the first time why the SeqSLAM algorithm is effective, and demonstrate the potential for cheap camera-based localization systems that function across extreme perceptual change.

Problem

• Evaluate SeqSLAM for localization in dark environments with blurred images
• Assess motion blur impact from long exposures during car movement
• Compare baseline image matching to SeqSLAM normalization techniques

Innovation

• SeqSLAM algorithm for dark environment localization
• Evaluates motion blur up to 10,000 ms exposure
• Patch and local neighborhood normalization analysis