Abstract: In order to content the development of the roboticized and intelligentized bus management，the management of self-service is adopted in many cities in China.But it can’t satisfy the request with bus adjusting and programing，so，it is a hot subject to develop a model of automatic passenger counting with high-precision to obtain the reliably daily flux of passengers.The precision of the present system is around 80%.The problem is caused by the crowed passengers when passengers take bus.The general pattern recognition is unable to distinguish sole person.Based on the sufficient research，the author design a project with common digital vidicon and the advanced model of P2DHMM.The result of the simulating experiment confirmed that it has nice capability on recognition，and affords a good foundation for the reality.

Key words: bus management, automatic counting system, Pseudo Two Dimensional Hidden Markov Model（P2DHMM）

摘要： 我国一些大中城市公交车运营管理采用了无人售票的刷卡系统管理，但是自动刷卡系统不能够满足车辆调度、运营规划等公共交通管理的要求，为了获取乘客流量的实际数据，开发高精度自动记数模型成为当前的热点和技术难点。我国现有乘客流量统计系统的精度在80%左右，造成精度问题的主要原因是在乘车高峰时段拥挤的状态背景条件下，一般模式识别算法和模型设计难以准确将单个人体识别和分割出来，为此，在充分调研的基础上，设计了采用普通数码摄像机和模板匹配方案，选择实验了先进的P2DHMM模型，模拟实验结果证实，该方法具有稳定的识别性能，为真实条件下的实验奠定了良好的基础。

关键词: 公交车运营管理, 自动记数系统, 伪二维隐马尔可夫模型（P2DHMM）