1） PC软件系统架构：摄像机+USB图像采集卡+PC电脑+RS485模块

2） 嵌入式系统架构： 摄像机+嵌入式系统平台

3） 火车速度<=40KM/H

4） 实现轨道左侧的信号灯颜色检测，红，白，蓝三色

5） PC软件接收RS485的信号进行相应的显示并通过485发送当前信号灯颜色

6） 嵌入式系统接收RS485的信号进行相应的显示并通过485发送当前信号灯颜色

7） 测试环境：雨、雾、冰雪等天气的白天和夜晚环境下

目前在PC上进行了初步的算法实验，演示视频见优酷（http://v.youku.com/v_show/id_XNTA3MzQ5ODA0.html）。大概的思路：首先检测出列出的2根轨道，得到大概的检测区域。并再用视频处理算法识别、跟踪信号灯。希望大家给些建议，使用哪种算法可以更好的跟踪信号灯。

-1) PC software system architecture: camera+USB image acquisition card+PC the computer+RS485 module 2) embedded system architectures: camera the+ embedded system platform 3) train speed < = 40KM/H 4) track lights on the left side of the color detection, red , white, and blue color 5) PC software receives the the RS485 signal corresponding display by 485 to send the current signal light colors) embedded system receives the RS485 signals corresponding display and signal light colors 7) by 485 to send the current test environment: weather such as rain, fog, snow and ice during the day and night environment on a PC preliminary experiments the algorithm, see demo video Youku (http://v.youku.com/v_show/id_XNTA3MzQ5ODA0.html). Approximate idea: first detected listed in the two tracks, and probably the detection region. And then the video processing algorithms to identify, track semaphores. We hope we can give some advice, which algorithm can better track the signal lights.