离散Hopfield神经网络在车牌识别系统中的应用

Application of Discrete Hopfield Neural Network in License Plate Recognition System

在一些特殊环境,如建设单位施工现场等,施工现场环境复杂,扬尘较大,采集的车牌汉字图像由于各种原因可能会出现变形、倾斜、污损、模糊和背光等情况,系统对车牌的识别精度明显下降。因此,提出一种基于离散Hopfield神经网络联想记忆的的车牌识别系统,在一定程度上去除了采集过程中出现的干扰。实验表明该方法具有较强的有效性和可行性,与传统算法在字符识别阶段加入深度学习的系统研究相比,该方法大大提高了车牌识别系统的正确率,提高了识别效率,优化了车牌识别系统。

In some special environments,such as construction sites of construction units,the environment on the construction sites is complex and the dust is relatively large.The images of license plate Chinese characters collected may be deformed,tilted,defaced,blurred and backlit for various reasons.The recognition accuracy of the license plate is obviously reduced.Therefore,a license plate recognition system based on the associative memory of discrete Hopfield neural network is proposed to remove the interference in the acquisition process to some extent.Experimental results show that this method has strong validity and feasibility.Compared with the traditional algorithm,which adds deep learning to the character recognition stage,this method greatly improves the accuracy of the license plate recognition system,improves the recognition efficiency,and optimizes the license plate recognition system.