一种基于视频的停车场车位监控算法

提出了一种基于视频的停车场车位监控算法。用户通过鼠标操作标定车位位置。监控算法基于以下三种判据：判据一是差影的均方值,可反映车位占用状态变化事件的发生;判据二是差影的方差,判据三是前景与背景比值的方差,通过判据二、判据三可以排除干扰、确认车位状态变化。当三个判据数值稳定时,更新车位背景。实验表明：该算法能够快速、准确地反映车位状态变化,易于实现。

Algorithm for Parking SPace Monitoring Based on Video

A vision-based algorithm for parking space monitoring is proposed.Users mark each parking area artificially by mouse clicks. The algorithm consists of 3 criterions: Criterion 1 is the mean square value of the difference image, and indicates occurrence of availability change of parking space; Criterion 2 is the variance of difference image; Criterion 3 calculates ratios of foreground to background at first, and then gets the variance of ratios. Criterion 2 and criterion 3 benefit the algorithm from avoiding interference, and therefore decide whether the availability of parking space has reversed actually. The algorithm periodically updates the background when values of all 3 criterions are steady. Results of experiments demonstrate that the algorithm performs well in processing periods and veracity rates, as well as practicability.