低分辨率条件下鞋类的自动分类方法

根据视频监控中行人所穿鞋的鞋型搜索犯罪嫌疑人是公安机关常用侦查技战法。然而在现实案件中很多视频监控分辨率较低,公安民警不能精确识别到具体鞋型,且消耗大量的时间和警力。针对这一问题,提出一种对低分辨率视频监控下的鞋类进行自动分类的方法。参考全国制鞋标准化技术委员会2017年制定的制鞋标准,初步将鞋类分为皮鞋和运动鞋两大类;构建鞋类数据库,包括59 853幅皮鞋和47 878幅运动鞋图像;进而基于卷积神经网络,设计一种适用于鞋类自动分类的鞋类识别网络模型。实验结果表明,鞋类自动分类模型在测试阶段对鞋分类的准确率达到了95.7%,可见基本能准确识别皮鞋和运动鞋两类鞋。

Automatic footwear classification under low resolution condition

In video monitoring, it is a common investigative technique for public security organs to search criminal suspects based on the shoe shape of pedestrians. However, in real cases, many video monitoring images are fuzzy, so that the police cannot accurately identify the shoe type, and it consumes a lot of time and police. In order to solve this problem, an automatic classification method of footwear under video monitoring is proposed. The reference standards set by the national technical committee on shoe making of standardization administration of China in 2017, the classification of footwear is systematically sorted out, and the shoe types are divided into two categories: leather shoes and sneakers. A footwear database was constructed, including 59,853 leather shoes and 47,878 sneakers; Then, based on the convolutional neural network, a footwear recognition network model suitable for automatic footwear classification is designed. The experimental results show that the accuracy of the automatic footwear classification model in the test stage is up to 95.7%, it can identify leather shoes and athletic shoes accurately.