A Real-Time Multi-scale Vehicle Detection and Tracking Approach for Smartphones
archive: archived pipeline: cataloged verified
Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)
Summary
This paper addresses the challenge of implementing real-time vehicle detection and tracking on smartphones, aiming to provide a low-cost alternative to expensive sensors like LiDAR for Advanced Driver Assistance Systems (ADAS). Motivated by the high prevalence of driver-related accidents and the increasing computational power of mobile devices, the authors present a vision-based pipeline integrated into the DriveSafe application. The primary goal is to overcome the strict computational constraints of smartphones while maintaining detection performance comparable to state-of-the-art methods. The proposed method utilizes a multi-scale detection approach combined with geometric road constraints to optimize processing efficiency. The system employs an AdaBoost classifier trained on Local Binary Patterns (LBP) features. Detection is divided into three stages: a low-resolution "near window" to identify close vehicles, a high-resolution "far window" focused on the vanishing point to detect distant vehicles, and an "intermediate window" that tracks previously detected vehicles to ensure continuity. To reduce computational load, the algorithm prunes the search space by removing the sky and vehicle bonnet, leveraging lane detection data to estimate the vanishing point. Tracking is managed using an Extended Kalman Filter (EKF) for position estimation and Optical Flow to handle motion between frames when detection fails. The algorithm was evaluated on the TME Motorway dataset, which includes challenging "Daylight" and "Sunset" subsets. Results demonstrate that the system achieves precision rates over 90% and recall rates exceeding 95% for vehicles within 60 meters, a distance range critical for driving safety. The method proved robust against significant lighting variations, performing similarly in sunset conditions. Comparative analysis showed that while precision was slightly lower than some specialized algorithms, the recall rate was higher, particularly for distant vehicles. Computational tests on iPhone 5 and iPhone 6 devices revealed average processing times of 132 ms and 76 ms per frame, respectively, allowing the system to run at 7.6 to 13.2 frames per second. This performance is sufficient for robust detection when operating at 5–10 fps, confirming the algorithm's viability for real-time execution on consumer smartphones. The significance of this work lies in demonstrating that effective vehicle detection and tracking can be achieved on affordable, widely available hardware without specialized sensors. By combining multi-scale detection with geometric priors, the authors provide a computationally efficient solution suitable for integration into broader ADAS frameworks. The results suggest that smartphone-based vision systems can serve as practical tools for monitoring driving behavior and enhancing road safety, offering a scalable alternative to complex, high-cost sensor suites.
Provenance
The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | Crossref | — | — | 1 | 2026-06-25 |
| archive | success | semantic_scholar | — | — | 6 | 2026-06-26 |
| extract | success | cached | — | — | 2 | 2026-06-26 |
| clean | success | clean | — | — | 1 | 2026-06-26 |
| chunk | success | chunk | — | — | 1 | 2026-06-26 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-06-26 |
| enrich | success | openalex | — | — | 1 | 2026-06-26 |
| promote | success | — | — | — | 1 | 2026-06-25 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 1 | 2026-06-26 |
| tag | success | vector_similarity | — | — | 6 | 2026-06-26 |
| verify | success | — | — | — | 1 | 2026-06-26 |
Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.
Topics
Ranked by relevance to this paper. Hover a topic for its definition.