毕业设计（论文）-交通标志图片分类工具的设计与实现.doc

盐城师范学院毕业设计 PAGE \* MERGEFORMATI 盐城师范学院毕业设计 PAGE \* MERGEFORMATI 盐 城 师 范 学 院 毕业设计 交通标志图片分类工具 的设计与实现 学生姓名 l 学 院 信息工程学院 l 专 业 软件工程 l 班 级 12（4） l 学 号 l 指导教师 l 2016年5月16日 交通标志图片分类工具的设计与实现 摘 要 现如今社会经济飞速发展，交通越来越便捷，道路交通的安全问题也日益受到人们的广泛关注。交通标志作为车辆行驶过程中信息传递的重要载体，已成为研究的热点。由于交通标志的种类多种多样且具有一定的识别难度，因此如何有效的解决交通标志图像分类问题成为当务之急。 基于以前的不同图像分类系统，本文采用更加适合于多种情况下交通标志的分类方法模型：卷积神经网络(Convolutional Neural Network, CNN)模型，以实现对交通标志图片的识别和分类。首先，我们需要定义卷积采样层，全连接层，逻辑回归层这三种层，然后将这些组件建立一个CNN模型。其次，要定义批量随机梯度下降(Minibatch Stochastic Gradient Descent, MSGD)算法的要素，包含成本函数的训练，验证和测试，参数更新规则2个要素，再运用MSGD来求解。系统最终实现了交通标志图片的分类。 【关键词】卷积神经网络；交通标志；图片分类；批量随机梯度下降算法 全套设计加扣 3012250582 Design and implementation of traffic sign image classification tools Abstract Nowadays, with the rapid development of society and economy, the traffic is more and more convenient, and the safety of road traffic is more and more concerned by people. As an important carrier of information transmission under driving situation, traffic sign has become a hot research topic. Due to the variety of traffic signs and certain recognition difficulty, it is urgent to effectively solve the problem of traffic sign image classification. Based on the different image classification system of the past, this paper uses more suitable method model for a variety of cases of traffic signs classification: convolution neural network model, in order to achieve the recognition and classification of traffic signs. Firstly, we need to define the three layers: the convolution sampling layer, the fully connected layer, and the logistic regression layer, and then build a CNN model by these components. Secondly, we need to define the elements of minibatch stochastic gradient descent algorithm, including the training of cost function, validation and testing. After the parameter update ruling two elements and we use MSGD to solve it. Finally, the syst