基于多目标优化多任务学习的端到端车牌识别方法

Multi-objective optimization based multi-task learning for end-to-end license plates recognition

中文关键词  车牌识别  多任务学习  多目标优化  深度神经网络  机器学习

英文关键词  license plate recognition  multi-task learning  multi-objective optimization  deep neural network  machine learning

中文摘要

      本文针对多个车牌识别任务之间存在竞争和冲突, 导致难以同时提高多个车牌的识别率的问题, 提出基于 多目标优化多任务学习的端到端车牌识别方法. 首先, 通过分析某些车牌识别任务容易占主导地位, 而其他任务无 法得到充分优化的问题, 建立基于多任务学习的车牌识别模型. 接着, 针对字符分割造成车牌识别准确率较低、鲁 棒性较差的问题, 提出基于多任务学习的端到端车牌识别方法. 最后, 针对多个车牌识别任务间难以权衡的问题, 提出一种基于多目标优化的多任务学习方法, 以提高多个车牌识别的准确率. 将本文所提方法在标准车牌数据集上 进行测试, 实验结果验证了该方法的有效性和优越性, 其他代表性方法相比可以提高车牌识别的准确率、快速性和 鲁棒性.

英文摘要

      In view of the competition and conflict among multiple license plate recognition tasks and the difficulty to improve the recognition rate of multiple license plates at the same time, a multi-objective optimization based multi-task learning for end-to-end car license plates recognition is studied in this paper. Firstly, by analyzing the difficulties that some license plate recognition tasks tend to dominate while other tasks cannot be fully optimized, a license plate recognition model based on multi-task learning is built. Then, aiming at the problem of low accuracy and poor robustness caused by character segmentation, an end-to-end license plate recognition method is put forward based on multi-task learning. Finally, a multi-task learning method based on multi-objective optimization is proposed to improve the accuracy of multiple license plate recognition. The proposed method is tested on the standard license plate datasets. Experimental results demonstrate the effectiveness and superiority of the proposed method, which can improve the accuracy, speed and robustness of license plate recognition compared with other representative methods.