As video compression improves and sensors sizes increase, the temptation to install hemispheric cameras grows. While there are some applications in which hemispheric cameras excel, it’s vital to understand the limitations and the alternatives, when you try to monitor 360-degree angles of view from a single location.
SOMETHING immutable is that 360-degree hemispheric cameras spread their pixels. Regardless of sensor size, this pixel spread will soften images, so you’ll get faces between 7-10 metres and situational awareness past that. As light falls, fine detail will be lost earlier, too. Depending on what you are trying to achieve, this softness in larger applications may mean no court-admissible face recognition. Hemispheric cameras offer one half of the surveillance puzzle – wide angle. But they can’t offer the other half – depth of field.
Alongside the serious compromise of pixel spread and shallow DoF comes a benefit. Situational awareness will be profound and at no time will a subject in the target area leave the field of view – you’ll be able to view events with complete context and continuity. Furthermore, hemispheric lenses have a small hyperfocal distance, so everything in the scene will be in focus. To make this sort of coverage work in court, you need your hemispheric view to be able to identify dress or general features of individuals you’ve already identified at an entry point using a longer lens with a much narrower angle of view – 8-10mm for a 1/3rd of an inch sensor.
There are other downsides. Any wide lens will be at risk of flare. Hemispheric CCTV cameras also have lashings of an optical aberration called barrel distortion, which bends straight lines and means that magnification decreases the further you go from the optical axis. With a hemispheric, the presence of barrel distortion is a characteristic that might be corrected but can’t be avoided.
Depending on your application, 360-degree cameras can deliver multiple image streams from one hemispheric camera – one will be the full view and around it will be arrayed images where the curvilinear scene is adjusted to a rectilinear format in-camera for more intuitive display. There will obviously be some loss of detail at the edges, top and bottom where pixel input is shorn off by software. Correcting software that relies on averages will also introduce softness to the final output.
There are alternatives to panoramic cameras when you are looking for complete coverage of an area – you can use multi-head cameras, each with a narrower angle of view, with the output stitched together by software in-camera. There are pros and cons to this approach. The sharpness of multi-head cameras is certainly better, but much depends on the purity of the software stitching – you’ll need to undertake objective testing to establish which is best for you. It may be that in more compact applications, a quality panoramic camera does a more integrated job at lower cost.
According to Mark Franklin, Sony’s technical product manager, the applications best suited to wide angle fixed lens CCTV cameras include situational awareness and overview applications such as foyer and reception areas, retail malls, stadia concourses, large floor coverage such as warehouses, and transport applications such as trains and buses. Franklin says that while such cameras offer plenty of benefits, there are also some pitfalls.
“Installing a small sensor, high resolution megapixel wide-angle camera and expecting it to cover multiple applications may not deliver consistent results across all applications,” Franklin explains. “For instance, there will be loss of detail at night or when relying on digital zoom achieve the desired angle of view. There may also be unnecessarily high bandwidths or images that do not maintain reliable colour rendition in varying light levels.
“So, it’s important to check requirements and then use required field of view and a lens calculator to choose a camera type that ensures these requirements are met,” he says. “Wide-angle camera can give good situational awareness and detail. For example, an FHD camera with an appropriate lens will maintain video standards and image performance day or night, and give good detail in a wide area application, such as foyer or reception areas. Meanwhile, 4K resolution cameras with an appropriate lens will maintain video standards at the same time, giving good detail in wide area applications such as malls, concourses, etc.”
A question is whether hemispheric cameras should always be used with higher resolution sensors or whether there are situations in which the combination of standard resolution and wide angle is a compromise worth making. According to Franklin, this depends entirely on the application and the use of images required.
“If the application is a foyer or reception entry area, then wide view and detail can be achieved with FHD (16×9 aspect) rather than by going to high resolution MP 4:3 aspect cameras,” he explains. “For example, FHD cameras with 3~9mm lens, when correctly installed, can identify a person/object while maintaining good dynamic range, low light sensitivity and good storage/bandwidth management.”
Another consideration with very wide-angle lenses is that there can be considerable lens distortion. There are ways to ensure that such distortions don’t impact on image quality but Franklin says that this barrel distortion is part of getting overall situational awareness from a single video stream using a hemispheric lens.
“Distortion correction can be used to combat distortion in wide angle cameras,” Franklin says. “For example, Sony’s 4K 1-inch sensor SNC-VM772R with its 8~25mm lens, (equivalent to a 3~9mm lens on a 1/3-inch sensor camera) employs this technique to successfully overcome lens distortion when set at its widest lens angle. However, when using hemispherical 360-degree cameras, de-warping techniques that present the original image as if it was from a standard lens camera will also leave some distortions, particularly at the edges of the image.
“Taking this into account, the benefit of a 360-degree situational awareness view from a single camera needs to be weighed up against the disadvantages. This decision will take many factors into account. For instance, recently introduced hemispherical 360-degree cameras, such as the SONY SNC-HMX70, also include on-board analytics, so that the camera is more than just an image sensor. It has the added benefit of being able to count people, objects, discern and alarm on colours, alarm on capacity and unattended object or object removed.”
So, what is the sweet spot in the viewing angle at 1080p for a typical depth of field – all sensor sizes being equal? Is a 100 to 120-degree angle of view as wide as you want to go at 1080p, or are there times 140 degrees is worth deploying?
“100 to 120-degree lenses provide a good angle of view for most wide-angle applications using 1080p 1/3-inch sensor cameras,” Franklin says. “Distortion correction is available in these typical view angles, but with wider lenses it is harder to achieve this and these applications, such as covering multiple turn-styles with a single camera at a stadium, may be better covered with panoramic and hemispherical camera rather than wider angle lenses to achieve better situational awareness.”
When you talk about wide angle views, you must think about multi-head cameras. A consideration for end users is that when thoughtfully installed their combination of 4x 1080p resolutions and 4 moderate wide-angle lenses may often be the best way to achieve 270 or 360-degree views without cabling multiple cameras. According to Franklin, this depends on the application and why you need the images.
“Multi-head cameras can assist in providing a good situational awareness provided that the pixel per metre and angle of view gives the same detail as a conventional camera,” he argues. “Camera theory still applies – these cameras should not be looked at as a way of providing a hang-and-bang, one-size-fits-all solution that’s really just a cost-cutting exercise.”
Installing a 360 multi-head camera in a way that assures an unhindered 360-degree view is challenging – how would you argue it’s best achieved?
“Both hemispherical and multi-head 360-degree cameras can provide exceptional situational awareness in open areas with un-hindered views, but with a multi-head camera, it also needs to be possible to independently adjust the separate sensors to work with obstructions. Obstructed views ultimately require a multiple camera solution and should not rely on a single 360-degree camera to cover the entire area.”
What’s the maximum range a typical panomorphic 360-degree camera can be expected to offer face recognition out to – would you agree it’s between 6-10 metres depending on sensor resolution and ambient light levels?
“The DORI principle (detection, observation, recognition, identification) applies so that the sensor resolution and the lens need to be taken into account when working out how many PPM at the target for a set distance,” Franklin explains. “Remember, a 5MP sensor in a 360-degree camera only has 2-3MP of usable image due to the hemispherical lens not utilising the whole sensor area. This limits the PPM at the target for recognition purposes when camera is mounted at the set distance.
“In comparison, a 360-degree camera with a sensor resolution of 12MP will give approximately 7MP (2640 x 2640) of usable pixels with a hemispherical lens and is going to result in better detail at the same distance from the subject. Higher resolution sensors then allow higher PPM to achieve recognition at these distances from the target. Larger sensor sizes improve detail and greater sensitivity in low light – for example, the Sony SNC-HMX70 with 12MP 1/2.3-inch EXMOR CMOS sensor.”
Are there applications where less extreme viewing angles deliver better quality images with more detail and less distortion?
“Yes, every video surveillance application is unique and should be assessed on its own requirements,” Franklin says. “There are many applications that can benefit from wide angle, multi-head or 360-degree panomorphic cameras but these should not be seen as a blanket solution. Our industry benefits from a wide range of camera solutions that should all be considered with respect to the use for which they were designed.
“Technology can provide tools to assist in getting the best out of each application, but ultimately it is important not to lose sight of what is required to meet the requirements of the images collected. This includes distance to target, target size, angle of view and sensor resolution required to achieve the required pixels per metre at target for the application.
“For instance, if a 360-degree camera requires facial identification for a set distance away from the target and this cannot be achieved because of loss of detail, then a conventional lens camera with correct lens/resolution combination may need to be used instead.”
According to Cecil Nie, product manager of Hikvision Oceania, the best applications for wide angle fixed lens cameras are are narrow spaces demanding an overall view, such as convenience stores, reception in hotels, meeting room or classrooms. He says that while there are benefits to these camera types, there are pitfalls to look out for.
“Wide-angle cameras bring larger coverage – this saves the cost of running multiple cameras, however, they can also generate distortion on the edge of the image,” Nie explains. “Due to the physical property of the wide-angle lens, the resolution of the edge of the image drops a lot compared with the middle angle ones. A higher resolution sensor would certainly offset the loss on the edge, so that we could have acceptable image quality of the full screen.
“Meanwhile, higher resolution provides the capability to identify the targets. Unfortunately, distortion is a corollary for getting one overall image of a large scene from a single lens. However, selecting a more professional wide-angle lens would strongly reduce the distortion and deliver a better image quality.”
Nie argues that 100-120 degrees is the best balance of angle of view and distortion with wide angle cameras but there are some applications where it’s worth going wider.
“While 100-120 degrees is good enough because it covers most applications with a minimum of distortion, there are times when a 140-degree angle of view combined with a resolution of 1080p could be the right solution for a small application as long as the end users can accept a certain level of distortion,” he says.
Would you agree that the combination of 1080p resolution and multiple moderate wide-angle lenses is often the best way to achieve 270 or 360-degree views without cabling multiple cameras?
“Yes, I would recommend it,” Nie agrees. “Integrated multi-headed cameras will not be a big problem for installations when the camera system provides a more natural image to fit human eyes by reducing image distortion.”
What about installing a 360-degree multi-head camera in a way that assures an unhindered 360-degree view – how would you argue that’s best achieved?
“You must consider the angle of each lens and make a perfect match out of them,” Nie explains. “It is not easy to make full use of the views of each lens in a multi-head camera unless the camera is pendant or ceiling-mount. As a result, installers usually prefer a lighter and smaller 360-degree multi-head camera. In some applications, a 180 to 270-degree camera would be a better choice since it offers more flexible installation, as well as large coverage.”
Kevin Saldanha of Pelco argues the ideal applications for cameras with wide angle views include building lobbies, the perimeters of buildings, parking lots and cost-sensitive applications that focus on camera counts and manned surveillance.
“A panoramic camera can replace a few fixed cameras and unlike a PTZ camera, it doesn’t need to be actively controlled by an operator,” he explains. “Users get great situational awareness at a glance because when installed correctly, panoramic cameras provide a seamless view across the entire field of view. Optimal implementations have excellent image quality that can provide detail in challenging and varying lighting conditions. Another advantage of panoramic cameras is that they capture everything without interruption and are never pointing the wrong way.
“Pitfalls include that fisheye cameras can vary in resolution from the centre of the lens to the periphery,” says Saldanha. “There are tiled multi-sensor implementations of panoramic cameras that provide multiple streams that require arranging and these can have gaps or overlaps. The best implementations of multi-sensor panoramic cameras, like Pelco’s Optera, are the stitch-and-blend images that present a seamless panorama an operator can zoom into for more detail with intuitive immersive views.”
Would you argue such cameras should always be used with higher resolution sensors or are there situations in which the combination of standard resolution and wide angle is a compromise worth making?
“Narrower field of view fixed cameras certainly have their place,” Saldanha explains. “They are great when you want to get detail on a targeted area. Users can put all the resolution available in the camera on that area of interest. You cannot get the detail at distance from any other camera than a long zoom PTZ camera which is great for reading licence plates from afar.
“But a panoramic camera is best suited to get a constant read across a wide field of view – situational awareness at a glance. These cameras complement each other nicely. With video analytics, we now have technology like Pelco’s Camera Link that can detect people moving in a wide area with Optera and have a nearby Spectra Enhanced PTZ camera also zoom in and track them.
“With fisheye lenses there is also variation from centre to edge in resolution that can be noticed at the edges of images. With multi-sensor panoramic cameras, there is uniformity of resolution across the lenses. De-warping algorithms eliminate the distortion and present immersive views naturally, much as you would see in a PTZ or fixed camera.”
When it comes to the sweet spot in the viewing angle at 1080p for a typical depth of field, Saldhana says cameras with 1080p resolution and 140-degree fields of view support applications where coverage of a large area is needed, fine detail is not required and low-cost is a high priority. Saldhana agrees that if thoughtfully installed their combination of 1080p resolution and multiple moderate wide-angle lenses can be the best way to achieve 270 or 360-degree views without cabling multiple cameras.
“Users also now have multi-sensor camera designs that have 180, 270, and 360-degree fields of views that present a single seamless view across the entire field of view,” he explains. “The Optera 270-degree model is a unique offering that’s ideal for the outside corners of buildings; users can see all the way around and directly below the camera for total coverage. Meanwhile, 180-degree models are great for wall mounting and covering sides of buildings while 360-degree cameras are often installed for coverage in parking lots.”
When it comes to installation techniques that deliver unhindered 360-degree views, Saldhana says that indoors, ceiling, or pendant-mounted 360 cameras present the best coverage.
“Outdoors, pendant-mounted 360-degree models have been used very successfully,” he explains. “However, users must account for a modest sliver of coverage loss from the pole they are mounted onto, which typically can be covered by an adjacent camera.”
When it comes to hemispheric cameras that combine panomorphic (fisheye) 360-degree lenses and higher resolution sensors, Saldhana argues these are great for coverage of wider indoor spaces like hotel foyers, school gymnasiums, and hallway intersections.
“Useful depth of field depends on the lensing and the resolution of the sensor,” he says. “Most camera models list these parameters on their spec sheets. In general, a fisheye or multi-sensor camera is best suited for situational awareness of a wide area. If you want face recognition of a person walking through a doorway, for anything but close distances, it is best to go with a fixed camera with adequate resolution and a field of view to match that entrance. If you are looking to read license plates at a distance, your best bet is a PTZ or fixed camera with a long zoom ratio.
“Every camera type has its strengths and is best suited for certain applications. Panoramic cameras fisheye or multi-sensor cameras provide situational awareness of wide areas. It is best to complement these with fixed IP and PTZ cameras to provide complete and directed coverage where needed so you can secure a given space.”