Automatic Integration of Ubiquitous Access Address in Camera Surveillance System Using Natural Language Processing

One of the issues rarely discussed in the IP camera is the system's difficulties in accessing the IP camera from different manufacturers. This problem is caused by the fact that each IP camera has an individual manufacturing preset access addresses that make the users unable to integrate IP cameras from other manufacturers automatically. It is complicated to access ubiquitous IP cameras automatically in an advanced surveillance system. This research proposes a selection of IP camera access links using the NLP extraction approach. TF-IDF is one of the NLP extraction techniques used in this research, which has been combined with a statistical approach as a novel method in order to select the reliable candidate address from the address data of the IP camera provided by the ISPY dataset. The experiment has been conducted using several schemes based on the percentage of degradation of the original access address. There are four proposed links results, which are 242 links (25%), 288 links (50%), 363 links (75%), and 388 links (100%). The evaluation using 10-fold validation in several simulations reveals that the proposed list can produce the range around 89% to 97% of average accuracy. Moreover, brute force complexity evaluation has denoted that the selected access address has a lower complexity than the original one.