基于多源数据融合的车辆误识别召回关键技术研究

Research on Key Technologies for Vehicle Misrecognition Recall Based on Multi-source Data Fusion

针对人工智能算法在车辆识别方面样本集影响较大的问题,提出一种基于多源数据融合的车辆误识别召回算法,首先采用深度学习算法对车辆图像进行第一阶段识别,应用车牌规则及登记数据区分识别结果,然后使用登记车辆、驾驶人员、卡口位置等数据,对识别错误的车辆图像进行再次分析,最后,与商业算法对比验证有效性。实验表明,车辆误识别召回算法改进深度学习算法依赖样本集的缺陷,识别准确率从76.78%提升到96.50%,召回率提高19.58%;污损车牌的识别准确率提高90%,期望为交通安全管理工作提供参考。

To address the significant impact of sample sets on artificial intelligence algorithms in vehicle recognition,a vehicle misrecognition recall algorithm based on multi-source data fusion is proposed.Firstly,a deep learning algorithm is used for the first stage of vehicle image recognition,applying license plate rules and registration data to differentiate recognition results.Then,data on registered vehicles,drivers,and checkpoint locations are utilized for a second-stage analysis of misrecognized vehicle images.Finally,the effectiveness of the proposed algorithm is validated by comparison with commercial algorithms.Experiments demonstrate that the vehicle misrecognition recall algorithm addresses the limitations of deep learning algorithms that rely heavily on sample sets,improving vehicle recognition accuracy from 76.78%to 96.50%and increasing the recall rate by 19.58%.Additionally,the recognition accuracy of damaged license plates improved by 90%,providing valuable insights for traffic safety management.