- This article is filed under:
- » View All
Security Comes Knocking on Machine Vision’s Door
by Winn Hardin, Contributing Editor - AIA Posted 08/10/2009
The ruggedness of machine vision hardware has attracted the attention of the security industry in recent years, but long distances between cameras and recorders, plus budget limitations, have kept security applications decidedly low-tech and away from machine vision advances.
Today, with the emergence of digital IP camera networks, video analytic software based on algorithms similar to machine vision inspection routines, and a willingness on the part of machine vision suppliers to alter designs to meet security needs, the security industry is taking renewed interest in machine vision products.
Machine Vision Helps Manage Secure Video
Typically, video security means analog CCTV cameras displaying images on video monitors in centralized control rooms. But with the advent of Ethernet-based, IP-addressable cost-effective machine vision cameras built for rugged environments, security customers are starting to realize the many benefits of digital imaging. These applications range from government-run monitoring of roadways, train stations, and airports, to privately-owned critical infrastructure security at chemical manufacturers, pipelines and nuclear power plants. But even in these ‘‘low tech’‘ imaging applications, machine vision is helping security integrators and end users with improved image compression and digital storage.
‘‘In security and surveillance applications, the need is to archive streams of video data using DVRs or hybrid systems,’‘ explains Keith Russell, President of machine vision supplier, Euresys (Itasca, IL). These applications typically require multiple cameras, the ability to switch with minimal delay between cameras (or simultaneous acquisition), and the best compression ratio using minimal bandwidth such as H.264. The majority of these applications still use analog cameras because of the long distances and existing infrastructure, and our frame grabbers are used because of high product reliability, accurate triggering of the camera, and our reputation for providing superior support. As mobile surveillance and ITS applications are located outside, our products must perform in a fairly extreme environment, and we do this very well.’‘
As ITS and surveillance applications grow in number, additional pressures are placed on the security personnel tasked to monitor a growing number of video feeds. The answer - and revolutionary event for the machine vision and security industries - is the evolution of video analytics and sensor fusion software. Like industrial machine vision software, video analytics uses image-processing algorithms to identify specific threats and even specific people. The right image processing software can tell the difference between a dog crossing a perimeter, or a person, and even identify the person using biometric queries to image databases.
Video Analytics: MV’s Way In
Custom camera and imaging system manufacturer, Imaging Solutions Group (Fairport, NY), has combined video analytics and smart cameras to help beat the bandwidth issues and subsequent data infrastructure costs facing airports and train stations as they use dozens - sometimes hundreds - of cameras to help identify potential criminals and terrorists in crowded public concourses.
‘‘We do a lot of airport and concourse security where you’re looking at a wide field of view at high resolution to read people’s faces,’‘ explains Kerry Van Iseghem, Founder of Imaging Solutions Group. ‘‘We detect the faces, run a facial recognition algorithm using our on-board Linux CPU and field programmable gate array, and then place that biometric information in the header of the database query file, which checks the face against a central database of known criminals and terrorists.’‘
Imaging Solutions Group’s facial recognition cameras for public areas typically include a 10-megapixel camera that can output at 1080p (30fps) and full resolution frame rates up to 12 fps. And while many would question the ability to run facial recognition biometric algorithms on a smart camera with a 600 MHz Linux CPU, Van Iseghem assures customers that this low cost processor combined with FPGA can do an incredible amount of processing cost effectively - particularly if it doesn’t have the huge computational overhead of today’s consumer operating systems - something you need on a PC host solution.
Biometrics is just one type of video analytic software, which generally involves automatic threat detection in still or streaming video. ‘‘The single biggest issue with video analytics is false alarms,’‘ explains Chris Ruttle, COO and CFO of Vumii Inc. (Atlanta, GA), a physical security company that specializes in actively illuminated near infrared imaging for identifying threats up to 2 miles away. ‘‘What people are moving to is a combination of video analytics with other sensors to reduce the probability of false alarms. This is another area where machine vision software may be able to help security.’‘
ISG’s high-spatial resolution and on-board processing capabilities fulfill two of the security industries most fervent wishes, but according to Vumii’s Ruttle there are numerous opportunities for machine vision suppliers to use hardware and software to position themselves for the security markets. Ruttle recently spoke at an AIA conference and had these suggestions:
- Better camera sensitivity while maintaining frame rates
- Stronger near infrared performance for both cameras and lenses
- Extended temperature ranges
- Longer focal length, higher zoom, higher aperture lenses
- Intelligence on board the camera capable of running multiple algorithms while giving end-users greater control over camera operation
- Affordable short-wave infrared cameras and emitters
- Affordable electron-multiplying CCD and CMOS cameras for low-light operations
- Megapixel resolution at higher frame rates while maintaining sensitivity.
- Sensor fusion software that aligns multiple video streams with other sensor feeds to give security personnel an intuitive view physical security conditions
Many of the Ruttle’s suggestions relate to digital imaging, which may seem counterintuitive based on the security industries propensity for analog systems. However, while analog cameras dominate the installed base in the security market, Ruttle says that today’s improved digital and IP-addressable cameras combined with improved flexibility in storage and transmission of video images is rapidly shifting the industry towards digital imaging solutions. Ruttle admits, however, that high resolution, high sensitivity, high frame rate cameras may be difficult to achieve at the price point the security industry wants. In the meantime, new government regulations are increasingly pushing industry towards adopting video security systems with greater capability than the standard CCTV.
‘‘The U.S. has been fairly calm since 9/11,’‘ explains Vumii’s Ruttle. ‘‘The general urgency for security upgrades has dried up, but we are seeing that government regulations are driving private companies to upgrade their systems. As an example, not only do chemical plants need to protect their perimeter, which they did with fencing and surveillance cameras, but they are expected to establish a protective buffer zone outside of the perimeter. That’s where we come in because a system like ours that can assess targets up to 2 miles away - fuse that with other sensor inputs and you optimize detection with fewer false alarms. We leverage a wide area system to initially detect potential threats so that our system can definitely identify them as friend or foe.’‘
Because of the multiple types of systems used to physically secure a location, and the high cost of serving and selling to government and private institutions, ISG’s Van Iseghem suggests the best way for machine vision companies to enter the security market is as a valued OEM supplier of cost-effective, unique solutions to larger security and defense contractors.
‘‘To sell directly to the state transportation and security agencies, you probably need $50 million just in marketing expenses,’‘ says Van Iseghem. ‘‘We also don’t want to compete directly with our customers. Our company is set up to serve OEM customers and we do it at a cost that’s identical to what the customer would spend to do it themselves. That’s why we’re successful.’‘
Both Vumii and ISG agree that the current economic conditions have delayed many security projects in the U.S., while overseas markets have remained strong because their security needs are more urgent and deployment schedules shorter than U.S. installations. Recently, however, projects delayed in the U.S. are starting to come to fruition. ‘‘Customers have been paying for a lot of non-recurring engineering; so many systems are not in full production yet. No one knows exactly how large the security market will be, but they are estimating large quantities.’‘
There are currently no comments for this article.
Leave a Comment:
All fields are required, but only your name and comment will be visible (email addresses are kept confidential). Comments are moderated and will not appear immediately. Please no link dropping, no keywords or domains as names; do not spam, and please do not advertise.