基于机器学习反馈的车辆自动路况识别

研究车辆路况自动识别的问题,提高识别的准确率和鲁棒性。针对车辆的路况自动识别系统极易受外界环境的影响,传统的基于PCA的路况识别方法在提取路况信息时无法避免恶劣天气等环境的影响,造成最终的识别不准确和鲁棒性不高的问题。为了克服这一难题,提出了基于机器学习的车辆路况自动识别系统。首先采用Haar小波特征提取方法,将受环境影响的路况图像中的有效特征准确提取并降维,然后利用支持向量机选择合适的特征参数,将特征参数输入到AdaBoost分类器中进行分类识别后就完成了最终的车辆路况自动识别,避免了传统方法自动识别受恶劣环境影响的问题。实验证明,这种方法能够有效克服外界环境的影响,准确完成车辆路况的自动识别,并且识别结果具有较好的鲁棒性和满意的效果。

Machine Learning Feedback Based Vehicle Automated Road Conditions Recognition

Research the problem of road vehicles automatic identification,and enhance the recognition accuracy and robustness.According to the traffic vehicle automatic identification system easily by external environment,the traditional PCA based road conditions identification method in extracting features cannot avoid the influences of bad weather and environment on traffic,causing inaccurate final recognition and low robustness.In order to overcome this problem,we put forward a machine learning based road vehicles automatic identification system.Firstly,Haar wavelet feature extraction method was used to extract the effective features in the the road condition images which were influenced by the environment and reduce the dimensions.By using support vector machine to choose the appropriate characteristic parameters,the feature parameters were input to AdaBoost classifier of classification after the final road vehicles completed automatic identification,avoiding the bad impact on the environment as in traditional automatic identification method.Experiments show that the method can effectively overcome the influence of the external environment and accurately complete vehicle traffic automatic identification,and the results has good robustness and satisfactory results were obtained.