基于改进BP算法的语音车牌识别系统

Voice License Plate Recognition System Based on Improved BP Algorithm

交通管理系统信息化、智能化的快速发展,对车牌识别的速度要求越来越高,性能更加完善。针对此,论文通过算法改进,增加语言播报功能,设计了基于Matlab仿真实现的语言车牌识别优化系统。该系统通过改进Otsu伪二值算法,快速的提升处理速度;基于行扫描灰度跳变分析的车牌定位,综合了基于纹理特征分析和基于边缘检测分析方法的特点,具有速度快和准确性高的优点;改进BP算法,提高网络训练的精度,神经元的训练速度,同时避免落入局部极小值点。实验结果证明,对随机拍摄的300幅车辆图像进行测试,识别准确率高于98%以上,系统能精准实现自然光环境下的车牌定位校正、分割和识别,识别结果通过语音播报,系统具有优秀的前瞻性和人机交互性。

Traffic management systems,intelligent rapid development of increasingly high demand for a more complete performance speed of the vehicle license plate recognition.For this improvement,through the algorithm to increase the language broadcast feature,design language license plate recognition based on Matlab simulation optimization system.The system to improve Otsu pseudo-binary algorithms to quickly enhance the processing speed,analysis based on line scan grayscale transitions license plate location,the method combines the characteristics based on texture feature analysis and analytical methods based on edge detection,with a fast and accurate the advantages of high,improved BP algorithm to improve the accuracy of network training,the training speed of the neuron,at the same time to avoid falling into local minima.300 vehicle images randomly taken test,recognition accuracy rate is higher than the 98%or more systems can be accurate under the natural light environment license plate location correction,segmentation and recognition,the recognition results through voice broadcast design system with excellent foresight and human-computer interaction.