行人重识别技术在洗煤厂中的应用研究

Research on Application of Person Re-identification in Coal Washing Plant

针对洗煤厂在行人重识别过程中由于设备遮挡、分辨率低、光线变化等原因造成行人身份识别为None及误识别的问题,基于FastReID算法对行人重识别技术进行优化。首先,设计了一种保存陌生人特征图的方法,以增加行人重识别技术的实用性;其次,根据初始行人身份信息,校正误识别行人身份信息,减少行人识别为None及误识别的次数;最后,优化人员特征图像库,补充行人特征图像,从而提高行人重识别的准确率。经实验表明,通过自动保存陌生人图像的方法在行人重识别过程中识别准确率可以达到91.87%,通过校正行人身份信息以及优化人员特征图像库的方法可以将行人身份识别为None次数及误识别次数均减少80%以上。

In order to solve the problem of pedestrian identificationis None and misidentification in the process of person re-identification in coal washing plant due to equipment occlusion,low resolution and light change,the person re-identification technology was optimized based on FastReID algorithm.Firstly,a method to save the stranger feature images is designed to improve the practicability of the person re-identification system.Secondly,according to the identity information of pedestrians,the misidentification of pedestrians is corrected to reduce the situation of pedestrian identificationis None and misidentification.Finally,the personnel feature image library is optimized to make up for the shortage of pedestrian feature information in the pedestrian feature image library,so as to enhanceperson re-identification accuracy.Experiments show that the recognition accuracy can reach 91.87%in the process of person re-identification by automatically saving stranger images,and the number of pedestrian identificationsis None and misidentification can be reduced by more than 80%by correcting pedestrian identity information and optimizing the personnel feature image library.