Vanishing point detection for visual surveillance systems in railway platform environments
针对铁路站台低分辨率监控图像，提出一种三步法检测消失点，通过多帧增强、自适应边缘检测和线聚类算法，在光照变化和振动下仍保持鲁棒，优于现有方法。
Visual surveillance is of paramount importance in public spaces and especially in train and metro platforms which are particularly susceptible to many types of crime from petty theft to terrorist activity. Image resolution of visual surveillance systems is limited by a trade-off between several requirements such as sensor and lens cost, transmission bandwidth and storage space. When image quality cannot be improved using high-resolution sensors, high-end lenses or IR illumination, the visual surveillance system may need to increase the resolving power of the images by software to provide accurate outputs such as, in our case, vanishing points (VPs). Despite having numerous applications in camera calibration, 3D reconstruction and threat detection, a general method for VP detection has remained elusive. Rather than attempting the infeasible task of VP detection in general scenes, this paper presents a novel method that is fine-tuned to work for railway station environments and is shown to outperform the state-of-the-art for that particular case. In this paper, we propose a three-stage approach to accurately detect the main lines and vanishing points in low-resolution images acquired by visual surveillance systems in indoor and outdoor railway platform environments. First, several frames are used to increase the resolving power through a multi-frame image enhancer. Second, an adaptive edge detection is performed and a novel line clustering algorithm is then applied to determine the parameters of the lines that converge at VPs; this is based on statistics of the detected lines and heuristics about the type of scene. Finally, vanishing points are computed via a voting system to optimize detection in an attempt to omit spurious lines. The proposed approach is very robust since it is not affected by ever-changing illumination and weather conditions of the scene, and it is immune to vibrations. Accurate and reliable vanishing point detection provides very valuable information, which can be used to aid camera calibration, automatic scene understanding, scene segmentation, semantic classification or augmented reality in platform environments.