Indoor Location Technology with High Accuracy Using Simple Visual Tags

Feng Gao, Jie Ma2023 · DOI

TL;DR

This paper introduces a novel indoor localization system that uses simple, low-cost visual tags to achieve high accuracy and robustness. By combining a fast tag-matching algorithm with a cooperative weighted least squares estimation strategy, the system effectively handles challenging scenarios like tag group singularity and environmental disturbances, outperforming traditional methods like UWB.

What Problem Does This Paper Address?

Accurate indoor localization is crucial for the growing field of intelligent mobile platforms, enabling navigation and autonomous functions. Unlike outdoor environments where GPS is readily available, indoor spaces suffer from signal attenuation and interference, rendering satellite-based systems ineffective. Existing indoor localization methods, such as those using Wi-Fi, Bluetooth, or Ultra-Wideband (UWB), often face challenges with high deployment costs, complex infrastructure, or susceptibility to multipath effects. Visual localization methods, like SLAM, can be computationally intensive and degrade in performance under varying environmental conditions. This paper aims to bridge this gap by developing a robust, accurate, and computationally efficient indoor localization system using simple visual tags, addressing the limitations of existing approaches.

Methods

The proposed system comprises two main components: simple visual tags and a high-accuracy location algorithm.

  • Simple Visual Tag Design: The tags are designed to be simple circles with distinct colors (RGB). This design choice was made because natural textures are difficult to control and easily affected by environmental factors, while coded tags (like QR codes) are more complex to detect and their texture can degrade. The simplicity of color and shape features reduces algorithm complexity and data storage requirements. However, this simplicity leads to a "nonunique problem" where tags might not be distinguishable by their features alone.
  • Fast Tag-Matching Algorithm: To address the nonunique tag problem, a fast query and matching algorithm is developed. It leverages the camera's Field of View (FOV) and the tag's azimuth. The system first estimates the camera's pose and position from the previous cycle using dead reckoning. This estimate, along with the camera's FOV model, helps narrow down the search space for potential candidate tags within the database. Candidate tags are then filtered based on their azimuth matching the observed tag's azimuth. This approach significantly reduces the search space compared to scanning the entire database.
  • Cooperative Pose and Position Estimation: The core of the localization algorithm involves estimating the camera's pose and position. The paper acknowledges that standard least squares methods can become singular or ill-conditioned, especially when tags are co-circular or their spatial distribution is unfavorable, leading to increased location errors. To overcome this, a Weighted Least Squares Method (WLSM) is employed. WLSM uses an error transfer characteristic analysis to assign weights based on the reliability of measurements. Crucially, a cooperative strategy is introduced that schedules WLSM with an interactive algorithm. This strategy allows the system to detect singular conditions and seamlessly switch to the interactive algorithm, ensuring robust performance even in challenging scenarios.

Datasets

Not discussed in the paper. The paper relies on simulations and experimental tests using a defined indoor environment.

Key Insights and Ideas

  • Simplicity is Key for Robustness: The paper demonstrates that using very simple visual features (color and shape) for tags, while seemingly limiting, can lead to a more robust system by reducing computational complexity and susceptibility to environmental changes like lighting variations.
  • FOV and Azimuth for Efficient Matching: Leveraging the camera's FOV and the tag's azimuth is a clever way to drastically reduce the search space for matching observed tags to stored ones, making the system faster and more efficient.
  • Cooperative Strategy for Singularity: The most significant insight is the development of a cooperative strategy that combines WLSM with an interactive algorithm. This allows the system to dynamically adapt to problematic geometric configurations (like co-circular tags) where standard methods fail, ensuring continuous and accurate localization.
  • Weighted Least Squares for Reliability: Using WLSM with weights derived from error analysis allows the system to prioritize more reliable measurements, effectively mitigating the impact of noisy or less accurate tag data.

Key Results

  • Reduced Location Error: Compared to the standard Least Squares Method (LSM), the proposed cooperative algorithm significantly reduces location errors, especially in challenging "cocircular" conditions.
  • Improved Accuracy over UWB: In experimental tests, the proposed visual tag system achieved an average location error of approximately 5.5 cm under disturbed environments, which is significantly better than UWB, with an average error of over 13.5 cm. The paper claims the location accuracy is higher than UWB by 62%.
  • Significant Error Reduction: The WLSM, when compared to LSM, reduced maximum errors in azimuth, longitudinal, and lateral location by 89.47%, 51.61%, and 88.78%, respectively, in simulation.
  • Robustness to Disturbances: The system demonstrated robustness to simulated environmental disturbances like light changes and random covering of tags.

Limitations and Future Work

  • Limitations:
    • The paper acknowledges that the simple tag features can lead to nonunique problems, which are addressed by the matching algorithm.
  • The performance of the system is dependent on the number and spatial distribution of visible tags.
  • While robust to some environmental changes, extreme conditions might still pose challenges.
  • Future Work:
    • The paper focuses on horizontal plane movement. Extending the system to 3D localization could be a future direction.
  • Investigating more complex tag designs or fusion with other sensor modalities could further enhance accuracy and robustness.
  • Exploring methods to further reduce computational load for even faster real-time applications.

Summary updated

BibTeX
@misc{gao2023indoor,
  author = {Feng Gao and Jie Ma},
  title = {{Indoor Location Technology with High Accuracy Using Simple Visual Tags}},
  year = {2023},
  doi = {10.3390/s23031597},
}
Figure