AUTOMATIC PEDESTRIAN CROSSING DETECTION AND IMPAIRMENT ANALYSIS BASED ON MOBILE MAPPING SYSTEM

Liu, X.; Zhang, Y.; Li, Q. · 2017 · DOAJ

DOI: 10.5194/isprs-annals-IV-2-W4-251-2017

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 need for automated monitoring of pedestrian crossings to ensure traffic safety and facilitate timely maintenance. As urbanization increases traffic density, pedestrian crossings suffer from wear and tear, leading to impaired visibility and potential safety hazards. The authors propose a system using vehicle-based Mobile Mapping Systems (MMS) to automatically detect pedestrian crossings and analyze their impairment status. This approach aims to provide low-cost, high-efficiency supervision of traffic infrastructure, aiding administrative departments in identifying defiled or impaired crossings for repair. The methodology employs a bipartite approach: automatic recognition followed by impairment analysis. For detection, the system uses images captured perpendicularly to the crossing direction. Initial recognition utilizes a Haar-like Adaboost cascade classifier trained on over 1,000 positive and nearly 5,000 negative samples. This stage prioritizes high recall to minimize missed targets, even at the cost of increased false positives. To refine these results, the system applies three filtering techniques: projection filtering to identify Areas of Interest (AOIs) based on horizontal belt distribution; contour information analysis to count crossing strips and distinguish them from other traffic lines; and monocular vision calculations to verify dimensions against standard specifications. For impairment analysis, the system extracts features from detection intensity histograms, such as total pixel count, peak values, and valid data width/height. These features are fed into an Artificial Neural Network (ANN) model to classify the condition of the crossing. The results demonstrate that the proposed algorithm achieves high recall, precision, and robustness across different lighting conditions and data sources. The projection filtering effectively eliminates most non-target objects, while the strip counting method successfully distinguishes crossings from other traffic marker lines, even when crossings are seriously defiled. The monocular vision verification further reduces false detections by comparing theoretical and actual image projections. The impairment analysis module successfully recognizes flaking paint and defilement, allowing for the automatic classification of crossing conditions. The system is capable of handling images from various sources, including MMS, provided the camera parameters are known. The significance of this work lies in its ability to automate the monitoring and maintenance of pedestrian crossings, reducing potential traffic safety risks. By providing a robust detection and analysis system, the approach facilitates the timely identification of impaired crossings, thereby safeguarding lives and property. The integration of computer vision techniques with MMS data offers a scalable solution for urban traffic infrastructure management, addressing the limitations of manual inspection and previous methods that were restricted to specific perspectives or unable to assess impairment. This contributes to the field of remote sensing and spatial information sciences by enhancing the automation of 3D road marking reconstruction and traffic facility monitoring.

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.

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.

• pedestrian behavior perception