Intelligent monitoring of indoor surveillance video based on deep learning

With the rapid development of information technology, video surveillance system has become a key part in the security and protection system of modern cities. Especially in prisons, surveillance cameras could be found almost everywhere. However, with the continuous expansion of the surveillance network, surveillance cameras not only bring convenience, but also produce a massive amount of monitoring data, which poses huge challenges to storage, analytics and retrieval. The smart monitoring system equipped with intelligent video analytics technology can monitor as well as pre-alarm abnormal events or behaviours, which is a hot research direction in the field of surveillance. This paper combines deep learning methods, using the state-of-the-art framework for instance segmentation, called Mask R-CNN, to train the fine-tuning network on our datasets, which can efficiently detect objects in a video image while simultaneously generating a high-quality segmentation mask for each instance. The experiment show that our network is simple to train and easy to generalize to other datasets, and the mask average precision is nearly up to 98.5% on our own datasets.

I. Introduction

Traditional video surveillance can only provide simple functions such as video capture and storage. It cannot automatically alarm for abnormal situations. In order to find abnormal behaviours in monitoring in real time, monitor personnel need to constantly observe the video. In this case, the monitor faces dozens of surveillance video images, which is easy to fatigue. It may not be able to respond in time to abnormal situations due to lack of concentration, and loss key information in the video. In addition, because a large amount of surveillance video needs to be stored for months or years, it will result in a large storage cost. Therefore, the intelligent video surveillance system is urgently needed to assist the monitoring personnel to use the intelligent detection technology to process, analyse and understand the video signal while retaining the original video key information, and automatically detect the target category and location information without manual intervention. In the event of an abnormality, an alarm is issued in time to effectively assist the monitoring personnel.