Page 4

EETE MAR 2014

The future of video surveillance is hyperspectral By Julien Happich Freze! You’re blushing, we see your blood pumping abnormally fast and our analytical software tells us that you contravene crowd flow statistics, which is yet another reason to closely inspect and match your 3D facial features to our would-be criminal database, that is, anyone we have on record for citizenship. Privacy advocate groups may manage to mute two-way audio surveillance cameras in some countries and the level of video coercion may not be felt so strongly by passers-by as most people are unaware of the level of analytics and geo-fencing that newly installed IPconnected HD surveillance cameras bring with them. But new surveillance cameras are no longer installed in a “record just-in-case then delete” scheme, in fact they are already adding their bulk of big data to the cloud. With Imec’s recent announcement that it now supplies hyperspectral imaging sensor technology to strategic partners for its deployment into commercial camera solutions including for global security markets, you can be sure this new capability will blend-in with the latest surveillance trends such as HD resolution and to some extent, stereoscopic vision. Of course, hyperspectral imaging is amazingly efficient at discriminating materials for sorting products, at identifying substances to check the freshness of foodstuff or to detect hazardous or illicit ones in airports. Implementing this technology into industrial machine vision or medical imaging can bring huge benefits to society. Medical applications are plentiful, from skin and tissue analysis for cancer detection A wandal-proof HD surveillance camera from Adimec. 3D One’s stereoscopic HD imaging modules. Tattile’s traffic light ANPR camera grabs up to 75fps with on-board analytics. to blood vessel imaging, or bacteria detection based on characteristic spectra. Because until now they were quite costly and bulky, hyperspectral imaging systems were mostly used in high-end remote sensing instruments such as satellites and airborne systems (for precision agriculture to assess crop quality or to identify different types of lands, contamination etc..). But imec’s breakthrough using narrow-band spectral filters at pixel level, applied through semiconductor thin-film processing, BaySpec’s OCI-1000 handheld hyperspectral imager. means that compact hyperspectral image sensors could be mass produced at low cost. This is a boon for camera vendors as it opens up new markets. Yet, where will this drive surveillance? How about performing real-time video analytics on your health status, churning data out of your skin’s spectral signature and combining this with crowd control tactics or even discriminating video geofencing? How about food and beverage vending machines spotting your blood-sugar level and making decisions (delivering a service or reversing to an “out of order” status) or reporting to insurance companies based on unstated health policies? Not anything you would notice from looking back at that seemingly innocuous lens. How much surveillance data is enough? The answer by authoritarian regimes is definitely “as much as technologically feasible”. It does not differ much from what large commercial companies would want when scrutinizing their customers, except sometimes they get blamed for their lack of transparency. And since it’s “for your own good”, why would you blame your government for being even more hungry for such data? The worldwide market for video surveillance equipment is expected to expand by more than 12 percent this year, according to market research firm IHS Technology, and any new feature that can help manufacturers differentiate from competition is good enough. Among Imec’s key partners for developing this technology is Adimec from The Netherlands, who offers to customize its cameras for defense & security imaging applications. Featuring a full HD (1920x1080 pixels at 30 or 60 frames per second) electro-optical sensor module, the company’s TMX-DHD series increases detection, recognition and identification capabilities, according to what is on their website. Another partner, 3D-One (The Netherlands) offers to couple high-resolution imaging and spectroscopy in its MO-50 and MO-60 series camera modules, combined with analysis and visualization software to extract the spectral information. The same company provides 3D (stereoscopic) HD imaging modules and analytical solutions to “accurately track people in a busy scene, such as a hall terminal or shopping mall” as it advertises on its website. Surely time will come when all these capabilities will be merged as a selling argument. Italian partner Tattile specializes in programming and developing Automatic Numberplate Recognition (ANPR) solutions, it offers mainly products for tolling, speed enforcement, traffic tracking and access control. Again, integrating hyperspectral imaging to its on-board embedded analytic solutions is probably on the agenda. Now the last partner cited is BaySpec (USA), who brought to market the OCI-1000, claimed to be the world’s first handheld hyperspectral imager. The palm-sized ‘point-and-shoot’ device is able to acquire full, continuous visible and near infrared hyperspectral data. 4 Electronic Engineering Times Europe March 2014 www.electronics-eetimes.com


EETE MAR 2014
To see the actual publication please follow the link above