You are Home   »   News   »   View Article

Video surveillance with behaviour recognition

Thursday, April 17, 2014

The latest development with surveillance video recording is behaviour recognition, when a computer tries to understand what the video image is showing. GlobaLogix explains how it installed a behaviour recognition video system for a US E&P company By JimFererro, Senior VP, GlobaLogix

Video cameras are the most visible and obvious aspect of physical security wherever anyone goes.

In that respect the oil and gas industry is no different in its application of cameras at remote well sites, facilities, communication tower locations, pipelines and at plants.

Given that pervasiveness, however, what do companies actually do with the cameras and the images they capture?

Three phases

There have been three phases in the evolution of how video is used for security - forensic, analytic and behaviour recognition.

The original use, and still the most common, is forensic. Images are captured and recorded to be viewed after the fact to see if the image captured can assist in prosecution. Some times, along with recording the image, video monitors are watched by security staff; this procedure is not effective or reasonable for multiple video channels.

The second phase in evolution was video analytic software, which is rule-based. Areas or virtual lines are defined in the image area that, if crossed by an object, sends alarms.

Unfortunately, this type of software is not effective for two reasons; it requires significant configuration for the rule definition and, secondly, the software is prone to false alarm problems. Although a virtual line can be defined, everything that crosses the line sounds an alarm. Not only a person or vehicle, but wildlife, vegetation (tumbleweeds), even shadows and light beams can trigger a false alarm, which requires additional configuration. Frequently, these systems are eventually shut off due to the sheer volume of false alarms and consequently the equipment/installations become a wasted expense.

The third phase in evolution has been developed by BRS Labs of Houston, Texas with its AISight software which employs behavior recognition. In other words, the cameras and the system become a self- learning machine that recognizes unusual events without configuration, which proves to be an exceptional advantage. Over a relatively short period of time the false alarms drop off significantly and the system only alarms on truly unusual states or events.

Case study - behaviour recognition

This new behavioral recognition software has proven itself in a video surveillance upgrade for one of the world's largest independent E&P companies.

The E&P company operates well sites throughout over 220,000 acres of core Eagle Ford Shale Play in South Texas.

Communication towers were in place as part of its SCADA and Local Area Network (LAN) systems. In Dewitt County, Texas alone, it has eight remote tower sites.

The provider, Houston-based GlobaLogix, installed Axis security cameras at these remote tower sites and installed video management system software.

On the project, several challenges were successfully conquered: ensuring the new software would actually decrease exposure to issues ranging from vandalism to safety and operations, overcoming the major shortfall about recognizing unusual behavior without requiring rules or custom programming, integrating robust video management system software into LAN resources without devouring limited bandwidth with video bursts and determining how and where alerts would be distributed.

The solution, through GlobaLogix's working relationship with BRS Labs, was to deploy BRS' AISight Behavioral Recognition Software, which advances situational awareness compared to rule-based analytics.

Machine-learning is the key. In the process of providing pinpoint accurate, real-time alerts to security personnel about genuine threats, the system continually 'learns' to ignore the trivial behaviors that trigger the flood of false alarms in other systems.

As a direct result, those tasked with important security concerns no longer must waste time with systems which 'cry wolf.' Machine-learning specifically means that through the system's artificial intelligence (AI) design it teaches itself which behaviors in a camera's field of view are normal within the security context and which are not.

The learning is instantaneous, beginning precisely when connected to a video network; the longer it operates, the higher quality the alerts become because of its ever-expanding memory bank.

AISight sends real-time alerts to safety/security staff upon recognizing abnormal behavior viewed by a camera. Subsequently it integrates with major video management systems and never requires tedious, non-productive reprogramming endemic to other systems. Additionally, since no rules must be created and maintained, the system can be deployed very quickly.

Results

The AISight solution proved its value by requiring less personnel involvement to set up, operate and maintain than other systems - in addition to delivering video surveillance without technological glitches. It enables flexible distribution of alerts, along with operational manageability based on overall volume and alerts relevance.

As a result, this oil company experiences lower exposure to issues including those of operations, safety, vandalism, downtime and unnecessary expense compared to the previous rule-based software.

This level of security is particularly helpful to organizations within this E&P company such as Physical Security, Operations and Health Safety Environment (HSE). They now receive actionable information from real-time alerts and, therefore, have a proactive awareness of abnormal events as they are unfolding in the vicinity of remote tower sites.

The elimination of costly and annoying false alarms is the real, overarching benefit of the system.

Oil and gas companies employing video now have a better option - behavior recognition. Machine-learning behavioral recognition software has become the new standard in surveillance by eliminating false alarms and allowing for real alarms that drive productive action.



Associated Companies
» Globalogix
comments powered by Disqus

CREATE A FREE MEMBERSHIP

To attend our free events, receive our newsletter, and receive the free colour Digital Energy Journal.

FEATURED VIDEO

Improving well head data by developing a well head reports Engineers Data Model
Nurhamizan Yussop
from Brunei Shell Petroleum Co Sdn Bhd

DIGITAL ENERGY JOURNAL

Latest Edition Oct-Dec 2017
Nov 2017

Download latest and back issues

COMPANIES SUPPORTING ONE OR MORE DIGITAL ENERGY JOURNAL EVENTS INCLUDE

Learn more about supporting Digital Energy Journal