PTZ Versus Fixed MP Camera Shootout

BE
Benros Emata
Published Apr 30, 2011 04:00 AM

As megapixel surveillance cameras become mainstream, many have begun questioning the value of PTZs. The main criticism of PTZs is that they can only provide good image quality by 'looking' at a very small area at any given time. Megapixel cameras, on the other hand, can see 'everywhere' at all times. With their significantly increased resolution, operators can simply digitally zoom to magnify any area of interest. With this approach, the megapixel camera provides the best of both worlds - capturing details at a distance while recording video across large areas.

But how true is this really? What drawbacks does a MP camera have compared to a PTZ and in what scenarios?

In this report, we share our test results of a shootout between a PTZ camera and 4 megapixel cameras (ranging from 720p to 5MP from Axis, IQinVision, Sony and StarDot). We wanted to see what tradeoffs would be revealed if we tested these cameras at the same time and in the same space.

We took our 5 camera test rig and set up at a moderately sized public parking lot. See the test location below:

We then set up a series of tests (during the day, twilight and nighttime) where a subject would move across the covered area. We tracked the subject with our PTZ camera and then contrasted the PTZ's resulting image with that of the megapixel cameras. The overview image below shows the dimensions of the test area and the key testing points we selected:

We then conducted competitive analysis for various scenes to rate and compare the relative performance of each camera. Here are the 15 scenarios we analyzed:

  • PTZ Capture Area vs MP Capture Area
  • Daytime Comparison by Distance / HFoV - 5MP vs PTZ
  • Nighttime Comparison by Distance / HFoV - 5MP vs PTZ
  • Midday Even Lighting (Far Subject)
  • Midday Even Lighting (Mid Subject)
  • Midday Even Lighting (Near Subject)
  • Twilight Harsh Lighting (Far Subject)
  • Twilight Harsh Lighting (Mid Subject)
  • Twilight Harsh Lighting (Near Subject)
  • Nighttime Lighting (Far Subject)
  • Nighttime Lighting (Mid Subject)
  • Nighttime Lighting (Near Subject)
  • Daytime License Plate Capture (Far Subject)
  • Daytime License Plate Capture (Near Subject)
  • Nighttime License Plate Capture

We found the results, and especially the tradeoffs, to be fascinating.

Megapixel cameras do not provide anywhere close to the level of range and detail that an operator controlled PTZ can provide. It is simply no contest. Claims to the contrary are little more than marketing hype. The level of disparity is dramatic. While the move from SD to MP has improved fixed camera coverage range, it is still nowhere close to what PTZs can provide. For instance, while the PTZ captured facial details at over 250 feet from the camera, the megapixel cameras struggled to capture even 30 feet. This disparity occurred even though we used a relatively narrow Fov (~50 Degrees) for the megapixel cameras. The relative results would have been far worse compared to a MP wide angle lens or a 360 degree camera (see our Theia wide angle test results and our Mobotix 360 test results for a comparison).

From our test results, here are our best cases for using megapixel vs PTZ:

  • The area to be covered is fairly small - say 300 square feet (or 30 square meters). This is tiny for outdoor areas but could be feasible in a small parking lot, entranceway, back alley, etc. In such areas, the MP can provide similar level of details for security purposes but with the added benefit of covering the whole area at all times.
  • The area to be covered is moderately sized - say a few thousand square feet (or a few hundred square meters) and you can accept getting mostly blurry images of people, cars and plates. The plus side is that the MP camera will provide relatively low quality images while the PTZ will provide nothing unless it happens to be pointed at a specific area where an incident is occurring.
  • You do not plan to have an operator monitor / control the PTZ. This case is a general argument for using fixed cameras over PTZs. Megapixel simply makes the fixed camera moderately better.

The video below overviews our key findings in a compact 10 minute video:

PTZ vs MP Capture Area

The image below provides a visual representation of the massive disparity between the PTZ's potential coverage area and the MP camera's actual coverage area achieved in the test. We defined coverage area as the geographical zone that the cameras can provide facial or license plate details.

Day Comparison By Distance / HFoV (5MP vs. PTZ)

The image below should help you understand the tradeoff between the two camera categories across a large area. Notice how the megapixel camera's image quality degrades substantially as the FoV widens but that the PTZ, using its optical zoom, can continue to provide a relatively detailed image. Remember, this comparison presumes the PTZ is actively tracking a subject.

Night Comparison By Distance / HFoV (5MP vs PTZ)

The image below shows the same contrast with the same fundamental tradeoffs, now at night:

Midday Even Lighting (Far Subject)

In this scenario, the MP cameras are digitally zoomed into our human subject ~275ft out where the horizontal FoV (HFoV) is ~256ft. The PTZ is maximally zoomed into the same human subject. The PTZ's HFoV at this distance is ~7.3ft. Note that the scene is slightly backlit due to the relative position of the light source or sun.

Below is the Composite Snapshot (Download the Full Video Clip [link no longer available]):

[link no longer available]

The MP cameras produce little detail at this range, with ppf ranging only between 5ppf (720p) and 10ppf (5MP). A human subject is detectable, but no identification is possible. In contrast, the PTZ zoomed maximally into the subject results in a fair level of detail, providing a significant increase in the likelihood of identification. Note that, due to backlighting effects, shadows across the subject's face has somewhat reduced the level of contrast and subsequent detail.

Midday Even Lighting (Mid Subject)

In this scenario, MP cameras are zoomed in to our subject ~130ft out, with a corresponding HFoV of ~121ft. The PTZ is zoomed into the same subject to an estimated ~5ft HFoV. Note that lighting conditions are even and not particularly challenging.

Below is the Composite Snapshot (Download the Full Video Clip [link no longer available]):

[link no longer available]

At this range none of the MP cameras are able to provide enough detail of the human subject for likely identification. This is not surprising, given that the ppf ranges at this HFoV is only between ~10.6ppf (720p) to ~21ppf (5MP). In comparison to the previous 'far' shot, however, we can begin to make out that our subject may have a shaved head, is wearing a black shirt and jeans, and we perhaps get a better sense of gender and build. The PTZ, on the other hand, at its advantageously narrow 6ft HFoV, provides considerable details, making identification highly likely.

Midday Even Lighting (Near Subject)

In this scenario, MP cameras are zoomed into the human subject located ~30ft out, with a corresponding 28ft HFoV. The PTZ is zoomed into the same subject to an estimated ~5ft HFoV, providing a calculated ~128ppf. Note that due to lighting conditions the scene is moderately backlit.

Below is the Composite Snapshot (Download the Full Video Clip [link no longer available]):

[link no longer available]

At this closer range the MP cameras are beginning to deliver a level of detail comparable to the PTZ camera. The 5MP (Axis P1347) and the 2MP (IQ042SI) are the best examples of this. The 1.3MP (Stardot) and 720p (Sony CH140) are not as 'crisp' as the other MP cameras, but still manage to reveal key details within the subject's face/body. The PTZ provides the highest level of detail among all the images in the composite, making identification highly likely. Note that, due to backlighting effects, shadows across the subject's face have somewhat reduced the level of contrast and subsequent detail.

Twilight Harsh Lighting (Far Subject)

In this scenario, the MP cameras are digitally zoomed into our human subject ~275ft out where the horizontal FoV (HFoV) is ~256ft. The PTZ is maximally zoomed into the same human subject. The PTZ's theoretical HFoV at this distance is ~7.3ft. Direct sunlight (~40K lux) into the cameras poses a considerable lighting challenge.

Below is the Composite Snapshot (Download the Full Video Clip [link no longer available]):

[link no longer available]

The harsh lighting of this scene produces ill-effects across all cameras (both MPs and PTZ), as evidenced in this composite by the slight 'haze' or washing-out now apparent. Further, the PTZ image has experienced a degradation in overall quality when compared to the less problematic 'Midday Even Lighting (Far Subject)' scenario. The majority of the MP cameras are now only providing silhouette or blob-like indications of the human subject as well. Of the MP cameras, the 720p (Sony CH140) is the strongest performer here, however, little beyond simple human recognition is possible given the severe lack of details.

Twilight Harsh Lighting (Mid Subject)

In this scenario, MP cameras are zoomed in to our subject ~130ft out, with a corresponding HFoV of ~121ft. The PTZ is zoomed into the same subject to an estimated ~5ft HFoV. Direct sunlight (~40K lux) into the cameras poses a considerable lighting challenge.

Below is the Composite Snapshot (Download the Full Video Clip [link no longer available]):

[link no longer available]

Again, harsh lighting in this scenario diminishes the clarity of images across all cameras (MPs and PTZ) when compared to the same HFoV during the 'Midday Even Lighting (Mid Subject)' conditions. Of the MP cameras, the 5MP (Axis P1347) provides the best level of detail. Not surprisingly the PTZ is providing a level of detail that makes identification highly likely.

Twilight Harsh Lighting (Near Subject)

In this scenario, MP cameras are zoomed into the human subject located ~30ft out, with a corresponding 28ft HFoV. The PTZ is zoomed into the same subject to an estimated ~5ft HFoV. Direct sunlight (~40K lux) into the cameras poses a considerable lighting challenge.

Below is the Composite Snapshot (Download the Full Video Clip [link no longer available]):

[link no longer available]

Even at this particularly close range the harsh lighting, again, produces an overall degradation in image quality across the entire composite (compared to its 'Midday Even Lighting' counterpart). However, MPs are still beginning to approach the clarity level of the PTZ in terms of providing identifying features/details of the human subject. High detail levels delivered by the PTZ in this scenario make identification highly likely.

Nighttime (Far Subject)

In this scenario, the MP cameras are digitally zoomed into our human subject ~275ft out where the horizontal FoV (HFoV) is ~256ft. The PTZ is maximally zoomed into the same human subject. The PTZ's theoretical HFoV at this distance is ~7.3ft. In this nighttime shot, ~3 lux of illumination is provided by artificial lighting in the parking lot.

Below is the Composite Snapshot (Download the Full Video Clip [link no longer available]):

[link no longer available]

It is immediately evident that the 2MP color-only (IQ042SI) provides no practical use under these circumstances (low light and long distance). In the rest of the D/N MP cameras, at least a human figure is arguably detectable within the image (perhaps only marginally in the 720p Sony CH140). The PTZ as well suffers a considerable loss in detail under these low-light circumstances, but still provides a fair enough level of detail to be of investigative use.

Nighttime (Mid Subject)

In this scenario, MP cameras are zoomed in to our subject ~130ft out, with a corresponding HFoV of ~121ft. The PTZ is zoomed into the same subject to an estimated ~5ft HFoV. In this nighttime scene, artificial lighting provides ~5 lux of illumination.

Below is the Composite Snapshot (Download the Full Video Clip [link no longer available]):

[link no longer available]

The negative impact of low-light is again evident across all cameras, especially the 2MP (IQ042SI) and 1.3MP (Stardot). Human detection is the most that can be determined, given the silhouette and blob-like appearance seen in the MP images. Of them the 5MP (Axis P1347) provides the highest likelihood of simple human detection. The PTZ image, though dark, clearly provides a much greater level of detail than the MP cameras. Even if identification is not a certainty in this case, several useful details can be obtained (shaved head, gender, build and facial hair).

Nighttime (Near Subject)

In this scenario, MP cameras are zoomed into the human subject located ~30ft out, with a corresponding 28ft HFoV. The PTZ is zoomed into the same subject to an estimated ~5ft HFoV. Note that artificial lights provide ~10 lux of illumination at this location within the parking lot.

Below is the Composite Snapshot (Download the Full Video Clip [link no longer available]):

[link no longer available]

In this 'near' scenario, substantial improvements in image quality across all cameras are observed when compared to the 'far' and 'mid' nighttime scenarios. The PTZ is providing a particularly high level of identifying details on the subject, while the 5MP (Axis P1347) and 720p (Sony CH140) begin to approach the PTZ's level of detail.

Daytime LPC (Far Vehicle)

In this scenario, the MP cameras are zoomed into a vehicle at ~220ft out, corresponding to a HFoV of ~205ft. The PTZ is zoomed into the vehicle as well, where the HFoV is an estimated ~6ft. The analysis focuses on the ability to read the vehicle's license plate. Note that intense sunlight directly into the cameras poses challenging lighting conditions.

Below is the Composite Snapshot (Download the Full Video Clip [link no longer available]):

[link no longer available]

The lighting conditions are quite severe at this particular location as suggested by the overview image (bottom right). None of the MP cameras provide any reliable identifying vehicle features (make, model, plate capture). In contrast, the PTZ 'shot' of the vehicle reveals enough details to identify the make (Toyota) and model (Camry). License plate capture is also possible to a fairly high degree of certainty.

Daytime LPC (Near Vehicle)

In this scenario, the MP cameras are zoomed into a vehicle located 68ft from the cameras. The corresponding HFoV is ~63ft. The PTZ is zoomed into the vehicle as well, where the HFoV is an estimated ~6ft. The analysis focuses on the ability to read the vehicle's license plate. In general, lighting conditions are challenging due to direct sunlight, but at this particular location lighting appears fairly even.

Below is the Composite Snapshot (Download the Full Video Clip [link no longer available]):

[link no longer available]

The 5MP (Axis P1347) provides the greatest level of detail amongst the MP cameras. Based on the 5MP's image the license plate characters can be 'guessed' with a fair degree of certainty. However, the PTZ's 'shot' of the plate provides capture to a nearly absolute level of certainty.

Nighttime LPC (Far Vehicle)

In this scenario, the MP cameras are zoomed into a vehicle ~127ft out, corresponding to a ~118ft HFoV. The PTZ is zoomed into the vehicle as well, where the HFoV is an estimated ~6ft. The analysis focuses on the ability to read the vehicle's license plate. Artificial lights provide roughly ~2 lux of illumination.

Below is the Composite Snapshot (Download the Full Video Clip [link no longer available]):

[link no longer available]

 

The PTZ's image of the license plate does not reveal enough details for an absolutely certain plate capture, but does provide enough for a fairly certain 'guess'. In contrast, none of the MP cameras provide enough to make even a remote 'guess' as to their characters.

Methodology

The following are the five (5) cameras used in the 'PTZ Versus MP Shootout':

  • Axis P1347 (online $1350) - 5MP D/N; 1/2.5" CMOS; F1.6 Kowa;0.08 Lux (B/W)
  • Canon VB-C60 (online $1225) - VGA D/N; 1/4" CCD; Built-in lens (F1.6 ~ F5.6); 0.2 Lux (B/W)
  • IQinVision IQ042SI (online $400) - 2MP Color; 1/3" CMOS; F1.6 IQinVision ;0.2 Lux (Color)
  • Sony CH140 (online $800) - 720p D/N; 1/3" CMOS; F1.2 Fujinon; 0.1 Lux (B/W)
  • Stardot SDH500BN (online $1010) - 5MP D/N; 1/2.5" CMOS; F1.6 Precision; 0.3 Lux (B/W)

Each of the five cameras were simultaneously recorded to an Exacqvision VMS during three (3) key lighting scenarios. All cameras are set to their defaults during test scenarios. All MP camera lenses were adjusted to a uniform lens angle (~50 degrees) and recorded/analyzed at their maximum resolutions. 

*Note that due to configuration issues with the Stardot SDH500BN its resolution was set to 1.3MP (rather than the maximum 5MP) throughout testing scenarios.

Here are the key default settings for each camera:

  • Axis P1347 - Exposure Value '50'; Exposure Control 'Automatic'
  • Canon VB-C60 - 1/30 shutter (fixed)
  • IQinVision IQ042SI -1/30 shutter; Color-only
  • Sony CH140 - 1/30s shutter; View-DR/VE On; AGC 'Middle'
  • Stardot SDH500BN - Extended Exposure 'On' (up to 1/20s); AGC 'On' (up to 64)

Here are the Three (3) key lighting scenarios we examined:

  • Midday Even Lighting - Even sunlight; ~10k to ~20k lux
  • Twilight Harsh Lighting - 30k - 40k lux directly into cameras
  • Nighttime - Artificial lighting varies from ~0.5 lux to ~10 lux

During each of the three (3) lighting scenarios a human test subject moves from a 'far' point (~275ft) towards the cameras. The subject makes momentary stops or pauses at various intermediate locations from throughout the inward transit. A 'mid' point stop/pause is made at a location ~130ft out and a 'near' stop/pause is made at ~30ft out. Several other pauses are made at close range as well (24ft, 19ft, 12ft, and 7ft). As the subject transits from 'far' to 'near' the PTZ is panned, tilted, and zoomed to maintain an optimal FoV on the subject (anywhere from 5' to 8' wide). When the subject enters within the 'near' point a PTZ preset is executed that corresponds to the same alignment/FoV of the MP cameras.