Shadow Detection Approach Combining Spectral and Geometrical Properties in Highway Video-Surveillance

Asaidi, Hakima; Aarab, Abdellah; Bellouki, Mohamed · 2012 · Crossref

DOI: 10.5120/8516-2564

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 cast shadow detection in highway video-surveillance systems, where shadows distort object shapes and interfere with tracking and recognition algorithms. The authors propose a generic approach that combines spectral, geometric, and temporal properties to automatically detect and extract shadows without relying on a priori assumptions about scene geometry or object shape. The methodology operates in two main stages: spatial analysis and color analysis. First, moving objects are identified using an adaptive background subtraction scheme, followed by noise removal and region merging via morphological operations. Edge detection is performed using the Sobel operator to identify object boundaries. The spatial analysis determines candidate shadow regions based on the direction of the shadow, which is calculated using the time of day, the site’s latitude, and the date. The daylight duration is divided into eight intervals, corresponding to eight possible shadow directions relative to the sun’s position. For each object, the extreme region in the determined shadow direction is selected as a candidate shadow region. The validity of these candidates is verified through color analysis. The approach assumes a linear mapping between the color values of a lit background region (antecedent region) and its corresponding shadowed region. By comparing the RGB intensity vectors of pixels in the candidate shadow region with reference pixels in the lit background, the system calculates a Euclidean distance. If the color coefficients are sufficiently similar, the region is classified as a shadow. This process integrates parameters related to illumination and surface reflectance to distinguish shadows from actual foreground objects. Experimental results on various test images demonstrate that the proposed approach is robust and efficient under different background conditions and changeable illumination. The method successfully detects shadows in real-world highway scenarios, confirming its ability to handle varied content and luminance changes. The significance of this work lies in its general applicability; by avoiding specific geometric assumptions, the algorithm can be deployed across a wide range of surveillance applications. The combination of temporal shadow direction estimation and linear color mapping provides a reliable mechanism for shadow elimination, thereby improving the performance of downstream video analysis tasks such as object extraction and tracking.

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.

• visibility analysis litigation