Abstract: In response to the problem of low traffic efficiency at intersections in the city due to the traditional rotation of traffic lights, this article proposes to rely on intersection camera image recognition technology and intelligently control the direction and duration of the intersection based on the dynamic data of vehicles queuing at each intersection. Considering that people want to queue as little as possible while not wanting their own queue length to be much longer than others' queue length, this article comprehensively considers minimizing the total queue length at intersections and minimizing the variance of queue length in each direction. The control strategy that minimizes the final result by weighting and summing the two factors is the optimal strategy. In order to compare with the alternate traffic strategy, this article uses computer generated simulation data based on field observations to calculate and compare, in order to achieve the optimal dynamic control strategy for intelligent traffic signals.

Key words: Image recognition; Intelligent traffic signal lights; Control strategy; Computer simulation

摘要： 针对现今市内各交通路口交通信号灯仅靠传统的轮流放行导致路口通行效率低下的问 题，文章提出依靠路口摄像头图像识别技术，根据各路口排队车辆动态数据，智能控制放行 路口方向以及时长。考虑到人们想尽可能少的排队，同时又不想自己排队长度远远大于别人 排队长度，文章综合考虑了路口排队总和最小化以及各方向排队长度方差最小化，将两个因 素加权求和，使最终结果最小化的控制策略即为最优策略。为了跟轮流通行策略做对比，文 章根据实地观察，用计算机生成相仿的模拟数据来计算对比，从而实现智能交通信号灯的最 优动态控制策略。

关键词: 图像识别；智能交通信号灯；控制策略；计算机模拟