The big image shows not only the width of this double image but the depth of field - it's impressive.

i-PRO WV-S4576L Panoramic Camera Review

♦ i-PRO WV-S4576L Panoramic Camera Review – i-PRO WV-S4576L Panoramic is a 12MP hemispheric camera that offers excellent situational awareness in a compact, robust form factor.

Distributed locally by BGW Technologies, i-PRO WV-S4576L has a ½-inch sensor delivering a resolution of 2992 x 2992 pixels at 30ips and there’s support from a 14-metre IR array. The squat 1.4mm lens is relatively slow, with an aperture of F1.9 – perhaps to give best possible depth of field per pixel.

The camera’s AI processor supports motion detection of people/vehicles, number counting, and congestion detection. It also supports third-party AI applications and meets a variety of AI demands. The data aggregated inside the camera can be displayed on a dashboard and used for marketing and business intelligence.

Minimum scene illumination is 0.3 lux in colour and 0.04 lux in monochrome, and 0lux with IR activated. WDR is 84dB, while image settings include automatic gain control and there’s adaptive black stretch, backlight compensation, fog compensation, highlight compensation and digital noise reduction.

The camera has an iVMD bundle, up to 8 privacy zones, a digital zoom offering 1, 2 and 4x and an angular field of view of 183 degrees horizontal and 183 degrees vertical. There’s Auto VIQs, microphone with G.726, G.711 and AAC-LC audio compression options, smart coding GOP control, 14 simultaneous users, and ONVIF profile options including G/S and T.

There are 3 alarm inputs, including input, VMD, Command alarm and alarm actions can be programmed to trigger SDXC/SDHC/SD memory recording, E-mail notification, HTTP alarm notification, indication on browser, and Panasonic alarm protocol output. The camera has a monitor output and a 3.5mm mini jack for audio input.

Power is 12V DC with a 1.04A current draw at approx. 12.5W, and the camera is also PoE (IEEE802.3af compliant) with a 12.5W draw.
Operating temperature is -40 to 60C, there’s a Rosahl dehumidification device, the camera is water and dust resistant to IP66(IEC60529), Type 4X(UL50), and NEMA 4X compliance, as well as shock resistant to 50J (IEC 60068-2-75 compliant), and IK10 (IEC 62262), with a wind resistance up to 150kmph at an ideal mounting height of 3 metres.

In the hand this a beautifully built surveillance camera, even without the base plate attached to increase heft. Dimensions when using the attachment plate are 154mm x 60.3mm (H) or 164mm x 96.3mm when using the base bracket. Weight is 880g without the base bracket and 1.3kg when using it. The camera is built to standards you’d expect from i-PRO in die-cast alloy and quality poly.

It’s rated IP66 against water and dust, IK10 against impact, it’s 50J compliant and has an integrated anti-condensation system. Outer screws are corrosion treated stainless steel and the dome is a clear polycarbonate resin.

Test Driving i-PRO WV-S4576L

We’re testing the camera using a TP-Link 5-port Gigabit switch with 4 ports of PoE connected to an HP TE9HSA0E Omen gaming laptop with an i7-8750 2.20GHz CPU, 165Hz QHD display, Nvidia GeForce RTX 3060 with Max-Q design (8GB GDDR6 and 16GB of RAM). The laptop is running Windows 11. Even though the camera is 12MP and we run at full resolution and full frame rate for 7 days, this machine had no issues with overheating.

In our application we’ve got the i-PRO WV-S4576L installed under the eaves on the corner of a building with wrap-around verandas. There’s almost no ambient light in this application after dark, so the camera is completely on its own in that regard. There’s also very strong backlight to the North from early morning onwards – up to 67,000 lux is measured by the Sekonic light meter at 11am. Because of the Northern set of the winter sun in Australia, there’s strong backlight/sidelight most the day. At times this provokes chromatic aberration around high contrast areas.

It’s immediately and endless apparent that the 360-degree angle of view misses nothing when it comes to situational awareness in our scene. The 360-degree hemispheric view gives the greatest detail, but the stacked panoramas are easiest from the point of view of live monitoring. Coverage is breathtaking – not only edge-to-edge but depth of field is comparatively strong, even though detail is inevitably impacted by pixel spread. It’s this overall breadth of coverage that keeps hitting home with the i-PRO panoramic during the extended period of our test.

Our rural application is not necessarily ideal for this i-PRO camera – there’s not a great deal of activity at our site, which is what we want in order to deliver analytics performance. However, what we do have here that we don’t have in our city application is a building corner. This hemispheric camera is simply perfect for a building corner.

The first image I see when the camera comes up onscreen is a ‘panomorphic ball’. This is a composite 360-degree fisheye of the entire scene. Sitting there resisting the temptation to spin the HP on its beams ends to view parts of the scene ‘right way up’, I go through all the display possibilities to see which one I like the most. What I choose is a dual panoramic where I see both sides of the house, including the corner and out onto the nearside of the lawn on both sides. This view rolls down the hill to the creek left side top, while the lower image shows another portion of the view – essentially everything back in the opposite direction.

It’s a very easy configuration to monitor intuitively and it’s such a comprehensive view it’s hard not to fixate on it. Because I have friends staying, I find myself using the views on my workstation to keep an eye on what’s going on around the place, simply because I can see so much. Discounting the walls behind the camera, I have nearly 270-degree angles of view to depth, including the entire lawn.

Looking down the hill, past the big tank and the old shed, I’m able to view all the way down into the corner of our fence and then down the hill to the neighbouring property, across the creek, and then over the lucerne flats to the line of the river 500 metres from the lens.

There’s a ford at the near creek and I decide I would be able to see farm vehicles on approach – it’s possible to discern plastic wrapped rounds of hay at much greater distance, so there’d be no great issue with seeing white farm vehicles at 250 metres, though you would not be able to identify make or model, just colour.

After living with the camera for a couple of days I discover that when it comes to detecting people on approach, you’d need to be watching closely to note movement at 70 metres, but even at 70 metres you definitely can discern intruders – as they get closer – from 40 metres in – you can make out details of attire, and establish what they are doing.

i-Pro has been making panoramic cameras for a while now and it certainly shows in this rectilinear image. The image is well constructed – image elements are not tortured by excessive distortion. There’s some softness in parts of the image when looking at the wide panoramas – in both cases it appears low down in the scene but too far from the lens to be an issue of hyperfocal distance.

What’s immediately and endlessly amazing about this camera is just how much of a view you get from a single camera point. Obviously, this observation applies to any panoramic camera, but in this large external application that’s the quality that really stands out. Our scene starts directly underneath and behind the camera in the sense that the rear quarter of the image is walls and veranda.

But once you get past the house, you’re then looking down into the valley – the home paddock fence line is nearly 100 metres from the lens and the line of the creek is maybe 250 metres, while the river is around 500 metres from the lens. Further still you’re able to discern elements of the hills behind. This deep situational awareness is soft but still useful and because of the angle of view, it’s a majestic unbroken sweep.

When it comes to identifying plates and faces, there are limitations of distance. With a 360-degree lens, even 12MP is going to suffer from pixel spread. While I discern softness deeper in with no target in the scene, once I wheel Norman outside and start moving him around, I find that past 10 metres I begin to lose the fine detail I’d have with a narrow angle lens. But there’s a major pay-off. Instead of face recognition to 15 metres across 100 degrees, I have court admissible face identification to 8 metres through 360 degrees.

Past the 8-metre mark, I have face recognition to around 10 metres through 360 degrees, past 10 metres I have varying degrees of situational awareness to considerable depth – this performance is best within 30 metres, but useful further in. Taken as a whole, these are extremely valuable characteristics for a CCTV camera to offer.

Day 1 of the test is a greyish day and it’s relatively late – around 3pm. The colour rendition is true – it’s not exaggerated. With all the rain there’s plenty of green in the grass but the browns are also accurate, and there’s considerable detail in the near wood grains, as well as in the rocks and structures in the angle of view. A fire has been lit in the field of view – the oranges and yellows are distinct. I note that little softness in the bottom of the bottom image and again in the bottom of the top image – it’s a double panorama remember – perhaps it’s a product of lens distortion at the edges – midsection detail is much sharper.

At about 4.30pm, the sun start sliding down behind the ridgeline to the North-West and amplification begins coming into play – there’s a little work on the image and the colour tone has altered slightly in line with the altered colour temperature of the available light – the scene has gone a little pinkish. Around about this time slight noise starts to push through the surface of the image. I move a car so it’s 8m from the lens to keep an eye on static plate ID in falling light.

At 4.40pm more noise is beginning to push through. Regardless, the image hangs on in colour – I’ve still got good colour rendition, including greens of foliage and the oranges and yellows of the fire – I can make out a person sitting in front of the fire at about 12 metres from the lens, but I don’t have identification or recognition. However, I can see the plate of the static vehicle at 8 metres.

Next time I sit down at my workstation around 5.44pm, the camera goes over into night mode and activates the IR array – this works best to around 10 metres from the lens in my application, though situational awareness is possible deeper in. Something I notice is that having that veranda beam so close is generating internal reflections in the lens, which would not happen in the field.

With the camera in night mode, I note detail of a person approaching within 5 metres of the lens – at this distance I have good detail and within 3 metres I’d argue a subject is near court admissible. Something else I notice is that when some ambient light is activated on the far side of the veranda the scene opens up, delivering more detail and greater sharpness.

I should point out a second time that this is a very dark application. There’s no ambient streetlighting out here, no light from the house next door, there’s not even a moon – I read zero lux on the Sekonic. Unsurprisingly, performance tails off out into the utter blackness where the IR array can’t reach. Closer in I have some situational awareness out there, but a panoramic camera is never as good at night as a narrow-ish angle bullet camera with a good-sized sensor and a powerful IR array.

Next day is sunny and I spend time trying to get a sense of how much information I can pull out of the body of my scene. Around 2.20pm there’s someone working down the hill about 35 metres from the lens. I’m getting hair colour, skin tone, clothing colour and I can tell the person is digging. I can see there’s a white vehicle down there – I can’t make out the type but it’s clearly a white station wagon or SUV.

And the key thing is that I have all that depth and level of detail – which gets sharper closer to the lens – throughout the entire angle of view. At 2.30pm I’m doing as well as I can in terms of things going on the scene – I’ve got a person in gumboots with light coloured pants in a maroon top and a fair complexion around 15 metres from the lens on the lower panorama, while on the top panorama I have the veranda, which includes the antics of a child and a dog. The child’s court admissible, the dog not so much.

Next time I look at the camera something has been hung from adjacent cabling well within the half metre hyperfocal range. This object is perfectly clear. By 3.25pm the light has changed, and the subjects continue moving around. I note a man and dog at 10 metres that’s near court admissible – certainly this person would be recognizable to those that knew them. A child in the top of the scene is at 6 metres I have court admissible recognition in this backlight, rather than identification. Regardless, performance is very good – there’s no trouble with hair colour clothing colour – and again, this performance is evident through the entire angle of view.

On this day I get a chance to go through day into night without interruptions. At 5.36pm the image has a pinkish orangey hue. It retains a good depth of field and reasonable colour rendition. There’s still that amazing panorama. I move the car around so I can get the plate from a different direction – you can nearly get the plate plates from a 7-metre distance, though moving plates are not really this camera’s forte. Situational awareness is this camera’s superpower.

It’s now 5.46pm and light is fading from the sky. The camera is hanging on and colour. Around 5.50p, I decide to zoom in on the car and I focus on a particular segment. This is a digital zoom so I’m not getting any more detail from the same number of pixels. At 5.51pm the camera goes into night mode, and wouldn’t you know it, in night mode I’ve got the plate.

 

 

I wind digital zoom back out and I’ve got the wide view with reasonable detail. As it gets darker, the scene closes in – I’m able to clarify the IR array is at its best in 7-8 metres from the lens. Next day I wake up early and take a look at backlight with sun just above the image – I can still see ice on the windscreen of the car and I’ve got the plate despite the backlight.

Norman Takes Centre Stage

At around about 1.15pm I wheel Norman out for a spin, starting at 3 metres from the lens. Close up like this I’ve got the ABCD lines, just as you’d expect. The afternoon goes along and the image stays strong. At 3pm I roll Norman all the way down to the other end of the veranda – around 8 metres from lens. I’ve got the D line and I’ve got separation on the C line but the B line and the A line have merged face recognition. I’ve also got face recognition and this test confirms what I thought I had when I was looking at a live subject the other day. I’ve also got his plate at that distance.

Next, I trundle him out to stand by the car at about 10 metres away from the lens. I’ve got the D line separated and the C line slightly separated, but it’s hard to be absolutely sure and there’s no clear ID but there’s nearly recognition. It’s the sort of picture that if you knew the person, you’d say yeah, that’s Norman. You can see his skin complexion, his haircut and colour swatches – the focus box has gone a bit Dr Who, but there’s still some detail.

Between 4.20pm and 4.46pm the light changes – I’ve almost got can see separation on the D and the C lines. I don’t feel that I can identify Norman so clearly in this light. I don’t see that the camera is putting work on the scene. At around 5pm I move Norman again – he’s 15 metres from the lens and I don’t have ID but I’ve got hair colour, I’ve got complexion. I can see separation of the D line. And I can see colours in his swatches – white, yellow, blue, red.

At 515pm I move Norman over the other side – he’s perhaps 10 metres from the lens. While I think it’s just the angle and there’s some reflection from him that I think I’ve got D-line separation but not C-line though his focus box is easy to see and I’ve got the swatches, too. Next, I shift Norman to 8m from the lens under the tree. I have the D-line separated but the C-line is beginning to merge, and A and B lines have merged. The swatches – I’ve got yellow, green, red and I think purple. I can see separation of the plate numerals.

What’s interesting in all this running around is that there could be 25 Normans spread through this enormous scene and I’d have the same level of detail of all of them. It’s 5.21pm and Norman is recognizable across 360 degrees. Finally, I move Norman a bit closer to a spot I intend to leave him for the night. At around 5 metres I’ve got the plate, the focus box, the swatches, D and C-line separation and going to tell myself I’ve got court admissible face identification in this image.

This pleasing image highlights what I’ve said before – 8 metres or closer is optimum for recognition. At 5.40pm light is fading fast but thanks to the angle of the setting sun I’ve got the focus box, plate, swatches, D line and C line. I don’t have identification, but I do have recognition. There’s some amplification pulsing through this image. It’s not too bad. And I still have good depth of field in all directions.

At 5.44pm the image starts to get a bit stressed but it’s not too noisy. I’ve got the focus box. I’ve got a recognizable plate on Norman. I’ve got the swatches. I’ve got the separated D and C lines – I have skin tone and hair colour. At 5.53pm the camera goes over into monochrome with IR and I have Norman’s plate. The focus box is not entirely separated, but maybe the first third in native resolution I’ve got D and C line separation. I can see the shape of Norman’s head and his hairstyle and even his facial ‘bone structure’, as well as vehicle make and model a few metres further back.

Conclusions

My experience with the i-PRO WV-S4576L panoramic camera was coloured by the flexibility and power of its 360-degree angle of view – aside from attaching it to a skyhook, a corner installation is almost the perfect application for this camera.

This i-PRO panoramic comes with a capable suite of analytics that’s extremely easy to pull into place with a mouse click. Analytics functions are colour coded, making the process of setting them up easier still.

Same as any hemispheric camera, faces and plates need to be within 10 metres or so for court admissible recognition, but pixel peeping into the scene is not really the point of a camera delivering such an astronomical level of situational awareness – especially one that has the video analytics to make the most of it. Instead, the i-PRO panoramic is like having eyes in the back of your head. Again and again during this long test I’m struck by the breadth of the image. It really is outstanding and I especially love the double panorama display.

If you’ve got a large angle of view over open ground with a little ambient light – a complex street or building corner applications with open ground on 2 sides – and you need recognition to 10 metres and absolute situational awareness much deeper into the scene, this i-PRO panoramic camera is an appealing choice.

Features of the i-PRO WV-S4576L:

* 12MP Sensor
* 2992 x 2992 pixel fisheye images up to 30fps
* Intelligent Auto (with AI Engine)
* Smart Coding (with AI Engine)
* ABF (Auto Back Focus)
* IP66, IK10, 50J compliant, Anti-Condensation System
* Built-in IR LED
* Fiber Optic Media Converter Unit (Optional WV-S25F1).

#SEN #SENnews #security #electronics