Vision Meets Wireless Positioning: Effective Person Re-identification with Recurrent Context Propagation

Abstract

Existing person re-identification methods rely on the visual sensor to capture the pedestrians. The image or video data from visual sensor inevitably suffers the occlusion and dramatic variations of pedestrian postures, which degrades the re-identification performance and further limits its application to the open environment. On the other hand, for most people, one of the most important carry-on items is the mobile phone, which can be sensed by WiFi and cellular networks in the form of a wireless positioning signal. Such signal is robust to the pedestrian occlusion and visual appearance change, but suffers some positioning error. In this work, we approach person re-identification with the sensing data from both vision and wireless positioning. To take advantage of such cross-modality cues, we propose a novel recurrent context propagation module that enables information to propagate between visual data and wireless positioning data and finally improves the matching accuracy. To evaluate our approach, we contribute a new Wireless Positioning Person Re-identification (WP-ReID) dataset. Extensive experiments are conducted and demonstrate the effectiveness of the proposed algorithm. Code will be released at https://github.com/yolomax/WP-ReID. CCS CONCEPTS• Computing methodologies → Matching.

Unsupervised Person Re-Identification with Wireless Positioning under Weak Scene Labeling

“…There are also some efforts [16], [17] to explore the use of multimodal data on person re-identification. Fan et al [16] utilize radar signals for person re-identification without considering visual data.…”

“…Fan et al [16] utilize radar signals for person re-identification without considering visual data. Liu et al [17] introduce the wireless positioning trajectories of mobile devices to assist crossdomain person re-identification. To measure the distance between the videos and the wireless trajectories, this method needs to manually label the GPS coordinate of each location under each camera scene to map the pixel coordinate to the world coordinate.…”

“…Fan et al [16] propose to use radar signals for supervised person reidentification, but additional radar equipment is required. Liu et al [17] introduce the wireless position trajectories for cross-domain person re-identification. Wireless positioning trajectories can be obtained through off-the-shelf equipment, such as cellular networks and WiFi positioning systems.…”

“…The immutability of signal ID allows it to assist vision-based person re-identification task. However, they need to label the GPS coordinate of each location of the monitoring areas to map the pixel coordinate to the world coordinate for calculating the distance between videos and wireless signals [17]. Such a paradigm suffers the scalability issue since in real scenarios, labeling and maintaining the location information of each position in all monitoring areas extremely consumes human labor.…”

“…Different from [16], our setting does not require new equipment. Compared with [17] which assumes the GPS coordinate labeling on the entire monitoring areas, our weak scene labeling setting only needs to access the locations of cameras. Therefore, our setting is more feasible and scalable for large-scale surveillance scenarios than previous settings [16], [17].…”