基于全局优化与深度学习的条形码识别方法

Approach of barcode recognition based on global optimization and deep learning

提出了一种在深度神经网络的基础上结合全局优化方法的条形码识别算法,利用卷积循环网络提取出条码中各字元的特征并进行分类,较传统方法具有更强的适应性与泛化能力,再进一步结合全局优化的方法,以达到充分利用条码结构性先验信息的目的,能显著提升方法的效果,尤其是将全局优化方法引入到神经网络中进行端到端学习,不仅保持了两者的优势同时还进一步提高了识别精度。实验结果证明了所提方法的有效性,达到约99.48%的识别精度,超越了传统的图像处理方法。

A barcode recognition algorithm based on deep neural network combined with global optimization method was proposed.The characteristics of each character in the barcode were extracted and classified by convolutional recurrent network.Compared with traditional methods,the approach is advanced in adaptability and generalization capability.By further integrating the global optimization to comprehensively make use of the barcode’s structural prior knowledge,the performance of the approach was promoted significantly.Particularly,importing the global optimization into the neural networks for end-to-end training not only maintains the advantages of both components,but also increases the precision.The experimental results prove the effectiveness of the proposed method,achieving a recognition accuracy of about 99.48%,which surpasses the traditional image processing methods.