Abstract
Blind individuals face persistent challenges in last-mile navigation, including locating entrances, identifying obstacles, and navigating complex or cluttered spaces. Although wearable cameras are increasingly used in assistive systems, there has been no systematic, vantage-focused comparison to guide their design. This paper addresses that gap through a two-part investigation. First, we surveyed ten experienced blind cane users, uncovering navigation strategies, pain points, and technology preferences. Participants stressed the importance of multi-sensory integration, destination-focused travel, and assistive tools that complement (rather than replace) the cane's tactile utility. Second, we conducted controlled data collection with a blind participant navigating five real-world environments using synchronized head- and cane-mounted cameras, isolating vantage placement as the primary variable. To assess how each vantage supports spatial perception, we evaluated SLAM performance (for localization and mapping) and NeRF-based 3D reconstruction (for downstream scene understanding). Head-mounted sensors delivered superior localization accuracy, while cane-mounted views offered broader ground-level coverage and richer environmental reconstructions. A combined (head+cane) configuration consistently outperformed both. These results highlight the complementary strengths of different sensor placements and offer actionable guidance for developing hybrid navigation aids that are perceptive, robust, and user-aligned.
Introduction
For blind and low-vision (BLV) individuals, the "last mile" of a journey—finding a specific building entrance, an empty seat, or a trash can—remains a significant hurdle. While traditional white canes provide excellent tactile feedback, they have a limited physical reach. Wearable computer vision, such as smartglasses, promises to extend this reach, but the optimal placement for these sensors remains a subject of debate. Head-mounted systems align with the user's gaze but often miss low-lying obstacles, while cane-mounted systems provide a ground-level view but are subject to extreme motion instability. This paper quantifies these trade-offs to guide the design of future assistive technologies.
Methodology and Experimental Setup
The researchers employed a two-part approach to investigate sensor placement. First, a qualitative survey of 10 experienced BLV cane users was conducted to understand current navigation strategies and technology preferences. Second, a quantitative evaluation was performed using Meta’s Project Aria smartglasses, which are equipped with five cameras and multiple Inertial Measurement Units (IMUs).
To isolate the variable of vantage point, the team performed synchronized data collection with one pair of glasses worn on the head and another mounted to a white cane. A blind co-author navigated five diverse environments, including outdoor sidewalks and complex indoor routes featuring elevators and doorways. The resulting data was processed using MultiSLAM for real-time localization and EgoLifter (a NeRF-based model) to create dense 3D reconstructions of the scenes.
Key Findings: The "Cane Paradox"
The study revealed a fundamental trade-off between localization stability and reconstruction richness, driven largely by the "Cane Paradox": the very sweeping motion that makes a cane useful for a human is highly disruptive for computer vision algorithms.
- SLAM Stability: Head-mounted cameras achieved over 98% pose accuracy across all locations due to the relative stability of the head. In contrast, cane-mounted accuracy dropped to 55–58% in outdoor settings. An ablation study showed that when the cane was held still ("No Sweep"), accuracy jumped to 83.2%, proving that algorithmic motion compensation is the primary technical hurdle for smart canes.
- Reconstruction Quality: Interestingly, while the cane performed poorly at localization, it often outperformed the head at reconstruction. The sweeping motion provides a high diversity of viewpoints that Neural Radiance Field (NeRF) models use to create high-fidelity 3D maps. Cane-mounted views captured ground-level geometry and low-lying objects that head-mounted cameras often overshot.
- Hybrid Superiority: The study found that a combined "Head+Cane" configuration consistently achieved the highest Peak Signal-to-Noise Ratio (PSNR) for 3D reconstructions, leveraging the strengths of both vantage points.
Conclusion and Future Work
The investigation highlights that head-mounted cameras are best suited for orientation and stable tracking ("landmark hunters"), while cane-mounted cameras excel as "detail scouts" for ground-level geometry. BLV users expressed a strong preference for tools that augment rather than replace the cane, valuing hands-free operation while recognizing the cane's unique tactile utility.
Future work aims to develop adaptive multisensor architectures that can dynamically switch between head and cane data—using the head for stable global positioning and the cane for fine-grained obstacle detection as the user approaches a target. This hybrid approach promises to create navigation aids that are both robust and closely aligned with the practical needs of blind users.
Summary updated