Nearly two years ago, I wrote a column reviewing the current state of panoramic cameras. Happily for users, the panoramic war has intensified and now represents one of the most exciting areas in video surveillance technology today.
For purposes of this column, “panoramic” is defined as a camera system capable of delivering continuous images over at least 180o when viewed with compatible software, as opposed to a multi-sensor platform delivering independent non-combinable tiled images. A camera’s performance is determined by sensor pixel count, lens and software.
There are two main approaches to panoramic cameras: single sensor and multi-sensor.
Single-Sensor Fisheye Cameras
Single sensor (“fisheye”) cameras have the edge when it comes to cost and simplicity. A single megapixel sensor is coupled with a fisheye lens to create an image of its surrounding area.
When mounted on a ceiling, a fisheye camera can provide 360° surveillance of the area below, sans donut hole. The 360° and 180° rectangular views are normally available for viewing. Displayed views are chosen via the video management system, which may also provide digital zoom and analytic functions.
“Fisheye cameras work best where you have completely wide-open areas, or in smaller areas where they give you everything in the scene with no blind spots,” explains Jumbi Edulbehram, Regional President of the Americas for Oncam Grandeye. “They also have the advantage of a flat profile and simplicity of installation.”
Two years ago, the sensors for these cameras fell into the 1.3 to 5 MP range, where Mobotix arguably was the market leader. Axis was in the mix, and Immervision lenses coupled with dewarping software created options from a number of other manufacturers. Today, that range has been extended with products from a number of other manufacturers, including Hanwha, Panasonic, Digital Watchdog (DW), and Oncam Grandeye, which offers a 12 MP camera line.
Pixel count in a fisheye camera goes beyond the sensor description listed in a data sheet. Published or highlighted specifications usually refer to the number of pixels on the sensor — a starting point. How many of these pixels will come into play when an image is presented to it depends on the lens.
For a uniform hemispherical lens, less than 60 percent of the pixels may actually see the scene. Lenses that are more elliptical in nature may get that number closer to 80 percent. Edulbehram sees square sensors coming into play in the future, upping the sensor’s pixel utilization.
The quality and consistency of lens manufacture is critical, since the dewarping software that will correct the image is tied to the lens, working as a system. Dewarping is essentially a flattening out of a spherical image, whereby the effects of geometric distortions are compensated for to a degree. Dewarping software works across the entire set of used pixels to create an image map, which simulates what the human eye would see directly or with a good conventional lens. In addition to optical compensation, dewarping must deal with the variance in pixel density near the scene edges. Not surprisingly, software is the key to the usefulness of a fisheye camera image.
Multi-sensor panoramic cameras typically use three or four imagers — each from 2 to 12 megapixels — to cover from 180° to 360°. Each sensor employs its own conventional lens, enhancing image uniformity and pixels on target. Further, depending on the camera, azimuth and rotation adjustments may be available to fine tune the image.
An exciting development is stitching software, which processes the images to combine them together in a continuous, seamless linear view. Optical spatial variances are not an issue in fisheye cameras, but compensation may be required for lighting variance produced by the sensors.
With higher pixels on target, really useful areas of interest, including the “stitched” boundary portions in the scene, can be identified and presented in separate video streams. Whereas the full camera image can create the situational awareness of its coverage area, an operator can select a portion for more detailed examination and use digital PTZ to bore in for closer examination.
Product offerings include Pelco’s Optera (4x3 MP sensors for 180°, 270° or 360° coverage) and DW’s new “Panzilla” (4x12 MP over 180°). Both products provide 12.5 and 15fps, respectively, at full resolution. “The race for resolution has ended — it is now a race for frame rate,” says Ian Johnston, DW’s CTO.
Which to Choose?
Kevin Saldanha, Senior Manager, IP Cameras at Pelco, emphasizes the importance of the “Panomersive” user experience (a Pelco term). “It is all about minimizing distractions and providing an engaging view for the user so they can quickly discern what’s happening across the scene,” he says. “With 120 dB WDR, to the stitching and blending of images, the panomersive experience delivers seamless panoramas and intuitive immersive views”
Ultimately, the planned use for the video should determine the camera type chosen. Is very high resolution going to be required for the application, or is lower resolution surveillance of a confined space more the requirement? Fisheye cameras are less costly (MSRP $500 to $1000) and provide a uniform blending lighting effect across a scene. Multi-sensor cameras may cost several times more, but deliver spatial fidelity and resolution.
While resolution and frame rate are key considerations, other decision factors include image quality, low light performance and WDR, analytics, diagnostics, cyber-hardening, stream characteristics, on-board storage, environmental hardening, and virtual PTZ presets. Some multi-sensor cameras may need a license per video stream, depending on the chosen VMS.
Beauty is in the eye of the beholder, so take the time to assess overall image presentation and features through your preferred VMS.
Ray Coulombe is Founder and Managing Director of SecuritySpecifiers.com and RepsForSecurity.com. Reach him at [email protected], through LinkedIn at www.linkedin.com/in/raycoulombe or on Twitter, @RayCoulombe.