DEVELOPMENT OF A PERSON RE-IDENTIFICATION METHOD BETWEEN MULTIPLE CAMERAS

Pedestrian's origin to destination traffic observation is expected to make use of video images. In this study, a classifier for person ReID was built by transfer learning using a metric learning method on the learned model OSNet. The accuracy of the classifiers was verified by capturing video images that observed at the real-world environments. We were able to propose a possible procedure for automatic observation of pedestrian ODs using multiclass classification.