Wifi MAC Address Tagging Assisted Fast Surveillance Video Retrieval System
Conventional public safety surveillance video camera systems required 24/7 monitoring of security officers with video wall display installed in the control room. When a crime or incident is reported, all the recorded surveillance video streams nearby the incident area are played back simultaneously...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English English |
| Published: |
2020
|
| Subjects: | |
| Online Access: | http://eprints.utem.edu.my/id/eprint/25453/1/Wifi%20Mac%20Address%20Tagging%20Assisted%20Fast%20Surveillance%20Video%20Retrieval%20System.pdf http://eprints.utem.edu.my/id/eprint/25453/2/Wifi%20Mac%20Address%20Tagging%20Assisted%20Fast%20Surveillance%20Video%20Retrieval%20System.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Conventional public safety surveillance video camera systems required 24/7 monitoring of security officers with video wall display installed in the control room. When a crime or incident is reported, all the recorded surveillance video streams nearby the incident area are played back simultaneously on video wall to help locate the target person. The security officers can fast forward the video playback to speed up the video search but it requires massive manpower if there are hundreds of video streams or multiple target persons required to be examined on the video wall. Even today with the Graphics Processing Unit (GPU) that is able to run the person search deep neural network model to automatic search for the target person from a large video database, it can take hours or even days to complete the search. This research aims to determine how to prioritize the surveillance camera video frames that need to be processed by the person search deep neural network model to reduce the time taken for getting the target person in the next camera (the cameras that may recorded the target person according to walkway topology). Thanks to the advancement in artificial intelligence, a person search deep neural network model trained to correctly match thousands of identical person can be used to automate the person search process. The person search matching process required the person in the image to be firstly detected before the matching can be carried out. Eight deep neural network based object detection models are re-trained on 55,272 labelled persons to determine the suitable object detection model that can be used to replace the person detection part of the person search model. As a result, applying Model 3 (Darkflow) for person detection is found to be able to provide reasonable speed/accuracy trade-off (0.62 mAP and 0.04s mean inference time). To further reduce the required time of automated person search without having to scale up the computing hardware, additional metadata (WiFi MAC address of smartphone) collected during the occurrence of the incident can be used to prioritize the retrieving of surveillance video frames for subsequent person search. Three ways of retrieving surveillance video are compared, in term of time taken for getting the target person, with a constructed testbed in UTeM. The developed WiFi sniffer enabled surveillance camera, with 3-stage WiFi frame inspection and the use of collected WiFi signal strength for filtering, is able to tag the collected WiFi MAC addresses to the surveillance video frames according to the time of the MAC address is sniffed. Using the formulated mathematical model, the proposed WiFi MAC address tagging assisted fast surveillance video retrieval method performs 9.6 times better in single person search and 6.2 times better in multiple persons search provided the WiFi MAC address of the target’s smartphone is sniffed by the WiFi sniffer of the surveillance camera. Based on these results, the proposed fast video retrieval system with MAC address tagging is proven to take less time to get target person in the next camera as compared to video retrieval system without MAC address tagging. Further research is needed to identify how to prioritize the WiFi MAC address searching when multiple WiFi MAC addresses are sniffed. |
|---|