Video Content Analytics on Media Search

Many large enterprises and organizations cumulated large volume of CCTV media resources in the form of images and videos which are typically stored in in-house databases with little or no access by the potential users after a period of time. There is a strong need to make these resources searchable at enterprise level besides using time.

The current media search engines rely mostly on time information, usually accompanied by human spending many man-hours in quick playing through to identify the subject of interest. They often miss relevant results as it is too time consuming to search the whole video.

One of the Singapore government agencies has accumulated over 500,000 hours of videos and still increasing. Therefore, it requires a centralized media management system to tackle and categorise the video clips by making use of latest Video Content Analytics (VCA) and Data Indexing technologies.

KAI Square Unified Platform with VCA

KAI Square Unified Platform with VCA is able to perform people and object counting, face recognition, license plate recognition, parameter defence and intrusion detection by using background filtering, colour search, people and object tracking and Optical Character Recognition (OCR) technologies. We have also adopted indexing and distributed processing methods to tackle numerous data at one time which rides on cloud computing.

How Indexing & Distributed Processing Help

When data (video clips) is being collected 24/7 from immigration customs and other public areas. Riding on KAI Square Unified Platform with VCA, all data (video clips) are stored in the cloud and processed by distributed computing which do not require any server by users. Collected data are distributed and parallel processed in cloud. Based on intelligent distributed computing, facial and time index processing can be performed in the centralized management system with minimum processing time. This has greatly saved time and manpower on categorizing the videos. Operators can search video clips based on time, colour and text.