Abstract: In order to calculate pedestrian flow from one station to another，this paper presents an automatic passenger counting technique based on passenger’s belongings extraction.By recording the figure of the passenger and identifying the color and location of their belongings，the method can extract and trace the passenger.According to the characteristics of bus passenger counting，this paper uses median filter algorithm，erosion and dilation algorithm，correlation analysis to extract passengers and identify their belongings.Also the paper uses partitioned processing method to simplify the process of calculation.By processing image sequences acquired from bus cameras，the experiment results show that this method works accuracy and efficiently.

Key words: automatic passenger counting, feature extraction, color-texture, partitioned processing

摘要： 提出了一种基于随身物品特征识别的自动乘客计数方法，其目的在于实现公交线路上任意两站间的客流统计。该算法的原理是通过提取乘客身高体型特征、随身物品的颜色位置特征来对乘客进行识别和跟踪。针对公交车自动乘客计数的特点，本算法引入特殊的中值滤波、腐蚀膨胀和相关性分析等操作，能有效地辨识出乘客目标及其随身物品信息。并通过分区域处理进一步简化运算。最后通过公交车视频的图像处理实验，证明该算法切实可行。

关键词: 自动乘客计数, 特征提取, 颜色纹理, 分区域处理