基于BP神经网络的轮胎花纹沟槽识别

李月芳，王希波^*，高岩飞，马飞燕，吕杭，刘广奇

山东交通学院 汽车工程学院，山东 济南  250357

摘要：针对传统轮胎花纹沟槽识别算法存在数据特征提取困难、数理运算步骤复杂等问题，基于BP神经网络对生成的不同沟槽类型的轮胎胎冠线数据集进行反复训练，得到BP神经网络轮胎花纹沟槽识别模型。将轮胎胎冠线数据集随机划分为训练集、验证集和测试集，通过试验验证BP神经网络识别模型对轮胎花纹沟槽的识别性能，由混淆矩阵得到模型的正确识别率为94.9%。从3、4沟槽轮胎中获取实际胎冠线样本数据测试BP神经网络识别模型的实际识别效果，6条胎冠线上的花纹沟槽数量全部识别正确。基于BP神经网络识别轮胎花纹沟槽数量具有可行性。

关键词：BP神经网络；轮胎胎冠线；轮胎花纹沟槽;识别

Groove recognition of tire tread based on BP neural network

LI Yuefang, WANG Xibo^*, GAO Yanfei, MA Feiyan, LÜ Hang, LIU Guangqi

School of Automotive Engineering, Shandong Jiaotong University, Jinan 250357, China

Abstract:In view of difficult in data feature extraction and complex in mathematical operation steps for traditional groove recognition algorithm of tire tread, the data set of tire crown line for different groove types is repeatedly trained, and the groove recognition model of tire tread based on BP neural network is obtained. The data set of tire crown line is randomly divided into three groups: training set, verification set and test set. The performance of the BP neural network recognition model is verified by experimentation, and the recognition rate of the model obtained from the confusion matrix is 94.9%. The actual sample data of tire crown line is obtained from the tires with 3 and 4 groove, which tests the actual recognition effect of the BP neural network recognition model. All the tire treads with 6 tire crown lines can be correctly identified. It is feasible to complete the groove automatic recognition of tire treads based on BP neural network.

Keywords:BP neural network; tire crown line; tire groove; recognition