CCTV Camera Performance: What Do You Need?
What is the right IP camera specification, balancing flexible form factor, angle of view, sensor size, resolution, etc, regardless of application, when seeking the best operational outcomes? Will a 5MP dome camera with 100-degree angle of view, H.265 compression, IP66 and IK10 rating, ½-inch sensor be ideal for most applications? Or is camera choice more challenging than that?
For end users and installers, camera choice is a key challenge in every application. Trying to balance opposing demands like face recognition in low light and low cost is an exercise in contradiction. A camera that did everything well for a reasonable price would make this choice easier, but no camera is ideal for every application. A camera that does well in low light is usually more expensive, A camera with a very low bitrate is often making compromises in the level of detail.
“It’s hard to say the best camera is X, as the situations are always so different,” explains Sargon Yousif of Axis Communications. “However, when it comes to our range, we can share that the Australian market still prefers a dome 1080P network camera. The AXIS P32 series dome has been our preferred camera for some time now – the AXIS P3245 is the fourth-generation model with the latest ARTPEC 7 processor with AI. In the last 18-24 months, there has been increase in the 4/5mp cameras and 4K, but still not enough to take first position. From the encoding perspective, H.264 seems to be still preferred as well, though more recently some installations are using H.265.”
Something that challenges many installers and integrators is that IP camera performance varies so much in areas like low light, motion blur and more. Brands can vary enormously, as can models within the same brand. Whether or not this comes down to the sensor, or the camera engine – including firmware – isn’t an easy question to answer.
“The engine has a lot to do with the process,” Yousif says. “The onboard processor is critical, but so is the lens and sensor. It’s the combination of the 3 that needs to be carefully selected to work together to achieve the desired outcome from the camera. Each camera will be different. For example, in the same series from AXIS P3245-LVE vs P3245-VE both outdoor 1080p cameras, their lens selection is different, as one will have IR and the other is designed to work without IR to produce colour video in low light.”
According to Axis’ Oshana Jouna, it can be important for users to get a sense of camera performance on their own sites before making decisions about cameras.
“The value of onsite testing depends on the use case,” he says. “If this is for a typical surveillance scenario, then testing camera onsite is not required. If the customer is comparing vendors and looking at datasheets to decide on which camera vendor to go with, then it is highly recommended to do a shootout.
“Surveillance imaging is different from broadcast imaging or imaging for other industries. In video surveillance it is all about trade-offs. IEC 62676-5 is attempting to standardize image quality for video surveillance but manufactures can still do the test to achieve the standard while compromising other parameters. For the customer to get a better understanding of camera performance a shootout is required (looking at the datasheet only is not enough).”
It’s fair to say that lens quality continues to drift downhill. What does Jouna think installers and end users can do to ensure they are getting the best value for money? What about lens speed – low light performance vs blooming for faster lens apertures under F1?
“At Axis we continue development with lens manufacturers to come up with technologies best suited to surveillance use,” Yousif says. “You have seen P-Iris lenses developed by Axis Communications and Kowa, and more recently I-CS lenses, again developed by Axis Communications and Computar, which the security industry has taken up well.
“We could spend hours on lens designs and differences. Lens speed or faster aperture is commonly used in photography, where typically you would want lower F-stop for the given focal length and field of view to improve low light performance. Lower F stop lenses need great aspherical glass to minimise light diffractions.
“If end users or installers are interested in the lens on the camera, look for the premium lens manufacturers like Fujinon, Kowa and Computar, especially when it comes to varifocal domes, which can be hidden or hard to see. Ask the camera manufacturer about the lens and why that lens was chosen for the sensor behind it. Do some tests in day/night with IR on and off to get a sense of performance.”
Another key question is whether the quality of H.265 compression compares favourably to H.264 compression? H.265 certainly has a lower bit rate but is there a quality cost to pay?
“This will highly depend on the scene,” says Jouna. “H.265 is primarily designed to reduce the storage needs of low-noise video with a lot of motion. In scenes where there is a lot of motion and low light, H.264 could have an advantage over H.265. The use of H.265 will also depend on the processing power available.
“The H.264 and H.265 standards do not stipulate the actual video compression method. Only the syntax and the method to perform playback is standardized. This enables improved video encoding solutions to be created while keeping the file format for interoperability (decoder compatibility).”
According to Jouna, Axis Zipstream technology is a more effective implementation of an H.264/H.265 video encoder for surveillance applications. It includes various surveillance-unique methods that enable networked cameras to produce video with significantly lower bitrate.
“Axis Zipstream technology provides both H.264 and H.265 support in parallel in the same products, without any need for reconfiguration or complicated system setup,” he explains. “True multi-streaming with per stream selectable codec and configuration enables both types of video to be transmitted or stored, for maximum flexibility. This twin-codec approach is central for making the transition period between the two standards as smooth as possible.”
If Jouna had to select the perfect camera for a low light application – where face ID unassisted was required – what would the specification be? What about if you needed plates?
“The ability to identify people depends on several factors,” Jouna says. “Some of the more important factors include camera resolution and scene size, lighting conditions, camera position, motion, and compression.
“Surveillance objectives determine the number of pixels a person or object needs to occupy in the captured footage. Axis recommends 80 pixels or more for identification in challenging conditions. For license plates, text should be 15 pixels vertically, so check legal requirements for footage intended as evidence.
“Identification might not be possible in challenging lighting conditions, even if resolution is high enough. Highly sensitive sensors and features such as wide dynamic range can help, but also consider better lighting and positioning of the camera to avoid backlit situations.”
Does the price pressure of CCTV cameras translate to operational performance, in your opinion, or are we at a point where camera performance is no longer the primary driver in CCTV?
“We are still seeing that camera performance is the main driver in certain industries,” says Yousif. “Especially end users who know what they are buying. There has been increased awareness on main processors being used, awareness around cybersecurity and greater focus on supply chains. Certain verticals still look for performance and quality as well – prisons, datacentres, transport and critical infrastructure are examples. They need to ensure that camera will last, deliver a quality image, and have ongoing firmware support and VMS integration support over the lifespan of the camera, while remaining cyber secure.
“For commercial/residential this may not be an issue as you can easily replace the camera if it dies within 2 years, but the labour cost of replacing a camera in a prison or tunnel can be 5 to 10 times the cost of camera itself.”
It’s fair to say that video analytics is becoming increasing important when it comes to quality CCTV cameras, but how many applications make best use of this technology and what are the most useful video analytics?
“Today’s video has amazing quality, amazing light sensitivity and that’s been driving the value of video for forensic purposes,” says Jouna. “Picture quality is so much higher today than it was only 5 years ago. On the other hand, analytics is still a work in progress, even though many of us think the industry is finally delivering what we promised 10 years ago.
“The most important analytic is the one that addresses the challenge presented by the customer. When looking at tender documents we often see a list of analytics that all cameras for a given project must comply with. This is regardless of the camera location or purpose. Generally, these requirements are not realistic, which results in analytics that produce poor outcomes with operators ignoring alarms.
“With the increase in processing power at the edge and the advancement in artificial intelligence (AI) we can make use of this technology and produce very accurate analytics at the edge, but the use case and the scenario has to be well defined for the analytics to work accurately,” Jouna says.
“Axis has launched cameras that are equipped with an embedded deep learning processing unit that supports third-party deep learning applications. This made it possible to run facial recognition and other processor intensive applications on the edge. It is now also possible to do object classification and use the info to do forensic search.
“With this in mind, consultants and end users should really study what analytics should be applied to which camera with a clearly defined outcome. This way the security integrator and manufacturer can supply and install the best camera for the particular application with the desired outcome.
“Still, I would envision analytics being more of a disruptor moving forward,” Jouna says. “It’s gaining some momentum – with deep learning, we’re able to make the algorithms even better and more accurate. Now we’re starting to provide some meaningful and accurate analytics that are useful for companies.”
Great detail in monochrome from this Axis dome.
Would Yousif agree the trend towards modest camera performance makes meaningful (court admissible) video analytics more challenging, at the same time as making searches much easier?
“While the ability of CCTV surveillance to reduce crime remains hotly contested in the literature, there is little doubt that it is often an extremely useful investigative tool,” he says. “Analytics have played a huge part in making it even easier to search for that few important seconds in a 24-hour period. Law enforcement is no longer reliant on witnesses who attempt to tell or reconstruct criminal events, often missing vital information. Investigators can collect visual evidence of incident in question from an unbiased source with instant and total recall of all.
“Unfortunately, there is very little data about the use of CCTV video in criminal matters before the courts, either overseas or in Australia. We have limited information about the way in which this footage was used by police and its impact on any later criminal proceedings. Despite the lack of data about its actual use by law enforcement in courts, media reports often highlight the importance of CCTV in criminal investigations. Often, we see on TV or online media which highlights the value of CCTV surveillance footage in several high-profile criminal investigations.”
Something that’s great about the latest CCTV cameras are flexible form factors – 180-degree angles of view, 360-degree views, robust bullets with decent focal length, modestly priced dome cameras that offer surprising low light performance, as well as IR support – do you think users and installers should delve deeper into the options in many applications, rather than using only one or 2 types?
“There has been significant improvement in the way the panoramic (180, 360 and multi-sensor cameras) have developed, and Axis is investing more in R&D in this area,” says Yousif. “We have not noticed our partners only use one or two models – they use a variety of models based on the application. More so from partners that use our presales engineering team, who can advise on best camera for the desired outcomes. There seems to be a preference towards domes in the Australian market, but street, city surveillance and retail surveillance are verticals in which panoramic or multisensor cameras are more commonly used.”
According to Andrew Cho of EOS Australia, lens, image sensor, and the processing chipset are the key elements of the performance with IP cameras.
“As most of the IP cameras nowadays are higher than 2MP, the quality in the cutting of the lens makes a significant difference,” Cho explains. “Then the size of the image sensor makes a remarkable difference as in low light applications it is essential for the sensor to absorb as much light as possible.
“However, the most important part is the processing chipset. The chipset processes the digital signal and turns into an image. Lastly, unlike the analogue cameras, IP cameras need to encode the raw image processed in the chipset, and the technique used for image compressing can make a noticeable difference.
Cho argues that there is an image quality differential when it comes to H.264 and H.265 and this needs to be factored into camera choice.
“When it comes to the quality of the video stream, H.264 compression is better,” he says. “However, H.265 compression value is much greater than the small compromise in image quality. The amount of the saving in the disk space, as well as the lightness in the transmission, makes H.265 is favourable to H.264.”
Cho says price does translate to operational performance when it comes to IP cameras.
“The price pressure has compromised the operational performance of CCTV in recent times,” he explains. “The low price and the higher resolution at a lower price tactic have blinded many security people into believing they are getting a better value without considering the real operational performance of the product. Not many realise a good quality 2MP camera has more useful information in its image than a cheap 6MP camera.”
How important is video analytics when it comes to quality CCTV cameras, in Cho’s opinion, and what are the most useful video analytics?
“Video analytics have opened up another horizon in the detection mechanism of CCTV systems,” Cho explains. “In the past, we relied on the motion event only, but thanks to the continuous development of video analytics, now we can utilise analytics to minimise the unnecessary storing of data, as well as improve efficiency in searching.
“Also, advanced analytics can proactively alert end-users rather being reactive. As an example, a loitering analytic allows users to be alerted and keep an eye on the suspicious movement. For example, a recent AI analytic from Hanwha enables the VMS to collect attributes for faster searching with detailed criteria, such as finding a male with glasses wearing a red t-shirt. These new AI analytics are becoming most useful.”
According to Cho, end users and their integrators should explore the possibilities offered by increasingly flexible camera form factors.
“Every camera design is developed with a thoughtful consideration of its usability,” Cho says. “The end-user will benefit not only with the performance/usability but financially as well. For example, the PNM-7000VD or PNM-9002VD from Hanwha, which has a dual-sensor in a single dome design, can easily replace 2 cameras with one. If the customer is using VMS, then the customer would only need to pay for 1 license instead of 2, and the cost to run the cable would be reduced.
“There are many other newly released products with optimum designs that would save on total cost of ownership. User and installers must delve deeper into available options to better design their security systems.”
Over at Hikvision Ryan Shi believes sensor and lens are the key contributors to camera image quality.
“Larger sensor sizes not only deliver better images, but also gives wider fields of view,” he explains. “Lens aperture has also became an increasingly important feature for cameras – a larger iris opening means more light coming through the lens to the sensor, giving better lowlight performance. Very high resolution used to be an important factor. However, with the development of camera technology and the market, a 4MP camera with an improved sensor and lens is adequate for most applications.
“When it comes to performance variations between cameras, sensors and lenses are the key contributors to camera image quality, especially low light performance. For motion blur, a powerful main chip further improves performance. Something else to consider is the settings of the camera for specific applications, things like frame rate and shutter speed are of equivalent importance for solving the problem of challenging environments.
“It’s very important to choose the right camera with better performance for industrial and critical applications. I would suggest consulting professionals for advice on the best cameras and the system design, ensuring the key requirements can be fulfilled.”
Installers and end users looking to ensure they are getting best value for money should focus on meeting their applications, according to Shi.
“For security surveillance, objectivity with clear image and target detail might be more important than fancy pictures with special effects,” he says. “The balance of cost and performance is also an important factor. In some ultra-low end markets, irrational competition might have caused some negative effect for the market. However, camera performance is improving over the years. Manufacturers are still striving to make better cameras to better meet the market requirement.
“When it comes to compression, H.265 has already become the prevalent option in the market compared to H.264 – H.265 codec saves a lot in terms of bandwidth and storage. Video analytics also helps improve efficiency – number plate, target classification and recognition, perimeter detection will be more popular and useful in the future. Given video analytics is based on image quality and detail, it’s better that camera performance and video analytics go hand in hand.”
According to Shi, flexible form factors – including cameras with 180-degree and 360-degree angles of view – are well worth considering.
“Absolutely yes,” he says. “Most end users and installers are familiar with the traditional camera types. They can still fulfil the needs and keep improving. However, new types such as panoramic and more-in-one models (multi-head cameras), would be better a fit for specific industries and niche markets, bringing more value and performance.”
At Hills, Anitha Bellary agrees that 2MP camera types remain a big seller, but she points out that the market is moving to higher resolutions, such as 5MP or 8MP.
“For the lens, a fixed focal length 2.8mm is our major volume driver, however, motorised varifocal lenses (2.8~13mm) dominate in projects, as these give greater flexibility of FoV,” she says. “H.265 is normal now for compression, while IP67 and IK10 ratings are for outdoor purposes only. The ideal sensor size is 1/2.8-inch, which is well balanced between cost and performance. True WDR is another basic feature, even if its performance varies between manufacturers.”
Bellary puts this variation between manufacturers down to a number of factors.
“It comes down to the lens quality, SoC and image sensor variations, the calibration knowhow/algorithm are also factors,” she says. “However, in most cases, performance is dictated by hardware and hardware-related engineering that optimizes performance.”
How important it is that users get a sense of camera performance on their own sites before making decisions about cameras depends on application, according to Bellary.
“If an application is unique or special, users’ sense of camera performance on the site is critical – for instance, users and installers should explore and implement new camera form factors to meet their applications,” she explains. “New form factors may provide great advantages to an application.
“Something installers and end users need to consider is that price pressure does translate to operational performance. That’s the reason we have multiple camera lines, so users have an option to choose based on their own operation, cost and level of acceptance of camera performance.”
When it comes to the choice between H.265 and H.264 compressions, Bellary argues that technically, image quality would be equal.
“Compression is the technology to reduce size of data while minimizing damage to image quality,” she explains. “More compression means more chance to damage the original image. However, due to the enhancement of image processing and CPU, you can barely tell the difference, even if size is decreased drastically (such as intelligent codec, Zipstream, H.265+).”
While Bellary agrees image quality feeds into the performance of video analytics, she says there are other factors at play.
“Edge VA cameras run VA processing and metadata generation at the camera end,” she explains. “Not only image quality, but also reliability of camera hardware and constant performance in various environments would be important. But again, without the association of a back-end application, its usage has to be limited.
“The use of analytics that shows clearest benefit is when it is replacing or supporting humans undertaking tasks like person matching through search of reference images and in COVID-19-related applications, where it is used for monitoring compliance to government guidelines.”