基于主成分分析网络的车牌检测方法

摘要

车牌识别是智能交通系统的核心技术,车牌检测是车牌识别技术中至关重要的一步.传统的车牌检测方法多利用浅层的人工特征,在复杂场景下的车牌检测率不高.基于主成分分析网络的车牌检测算法,能够无监督地逐级提取车牌深层特征,可有效提高算法的鲁棒性.算法首先采用Sobel算子边缘检测和边缘对称性分析获取车牌候选区域;然后将候选区域输入到主成分分析网络中进行车牌深度特征提取,并利用支持向量机实现对车牌区域的判别;最后采用非极大值抑制算法标记最佳车牌检测区域.利用收集的复杂场景下的车辆图像对所提方法的参数进行分析,并将其与传统方法进行比较.实验结果表明,所提算法的鲁棒性高,性能优于传统的车牌检测方法.%License plate recognition is the core technology of intelligent transportation system(ITS).License plate de-tection is a crucial step in the license plate recognition technology.Since only low-level artificial features are used to achieve license plates detection in most traditional methods,the detection error rates are usually low in complex scenes. In this paper,a novel license plate detection method based on principal component analysis network(PCANet)was pro-posed.Firstly,the license plate candidate area is marked with Sobel operator based edge detection and edges symmetry analysis.Secondly,by inputting candidate area into PCANet,the deep feature extraction is peformed for candidate area in PCANet and the support vector mechine is used to confirm the license plate.Finally,an efficient non maximum sup-pression(NMS)is used to label the best license plate detection area.For performance evaluation,a dataset consisting of images in various scenes was used to test the proposed method,and the results were also compared with those of tradi-tional methods.The experimental results show the robustness of the proposed algorithm,and its performance is also su-perior to the traditional method of license plate detection.