Big image - first guy about 6m from lens, second is 14 metres, third 16 metres - road workers at least 25 metres.

Hikvision 4MP Ultra Dome Review

♦ Hikvision 4MP Ultra Dome Review – Hikvision’s DS-2CD3747G2-LZSU 4MP Ultra Series ColourVu dome camera features a 1/1.8-inch progressive scan CMOS sensor and delivers colour images in 0.0005 lux at F1.0 with AGC on, or 0 lux with integrated warm/white light activated.

Ultra Dome has a fast F1.0 3.2-9mm varifocal lens with an MgF2 coating on the front element, employs H.265, H.264, H.265+, H.264+ compression options (I use default H.264) with various options across 4 streams, has an IP67 rating against water and IK10 against vandalism, hustles 130dB of WDR, has 40m of integrated white light, audio and alarm interfaces, and does neat things like human and vehicle classification based on deep learning.

The sensor delivers a maximum resolution of 2688 x 1520 pixels, there’s CBR/VBR scalable video coding (SVC) and H.265 and H.264 encoding, as well as 5 fixed regions for main-stream and sub-stream. Meanwhile the lens has a horizontal field of view between 92.3 and 48 degree, and a vertical field of view between 48.4 and 27 degrees, while the mechanism pans from 0-355 degrees and tilts 0-75 degrees and the integrated white LED light has a range of up to 40 metres.

DORI numbers are: Detect from 69-124 metres, observe from 27.4 to 49.3 metres, recognise from 13.8 to 24.8 metres, and identify from 6.9 to 12.4 metres. While we were not able to detect extremes of performance at the long end, these numbers were borne out by performance – particularly with recognise and identify to 24.8 metres. Often DORI claims are a little optimistic but these are very close to the mark.

Camera image settings include rotate mode, saturation, brightness, contrast, sharpness, gain, white balance adjustable by client software or web browser, there’s day/night function, 52dB signal to noise ratio and image enhancement that include BLC, HLC and 3D-DNR.

Hikvision 4MP Ultra Dome Review 2

When it comes to edge analytics, the camera offers motion detection (human and vehicle targets classification), video tampering alarm, exception, and a smart event function for scene change detection and audio exception detection.

Deep learning functions include face capture and perimeter protection supports line crossing detection, intrusion detection, region entrance detection, region exiting detection, unattended baggage detection and object removal detection.

The camera has an RJ45 port, a 10 M/100 self-adaptive ethernet port, on-board memory storage for microSD, SDHC, and SDXC up to 256GB, a built-in microphone, an alarm input/output, and audio compression options including G.711ulaw, G.711alaw, G.722.1, G.726, MP2L2, PCM, MP3 and AAC, and environment noise filtering.

Simultaneous live viewing is up to 6 channels with up to 32 registered users in 3 user levels – administrator, operator and user. Network protocols supported are extensive and include TCP/IP, ICMP, HTTP, HTTPS, FTP, DHCP, DNS, DDNS, RTP, RTSP, NTP, UPnP, SMTP, IGMP, 802.1X, QoS, IPv4, IPv6, UDP, Bonjour, SSL/TLS, PPPoE, SNMP, and ARP. API options include Open Network Video Interface (PROFILE S, PROFILE G, PROFILE T), ISAPI and SDK.

When it comes to security, there’s password protection, complicated password, HTTPS encryption, IP address filter, security audit Log, basic and digest authentication for HTTP/HTTPS, TLS 1.1/1.2, WSSE and digest authentication for Open Network Video Interface. Meanwhile, network storage options include NAS (NFS, SMB/CIFS) and auto network replenishment (ANR). Hikvision memory card, memory card encryption and health detection are also supported. The camera client options are iVMS-4200, Hik-Connect and Hik-Central Web Browser.

Hikvision Ultra Dome is nicely built of cast alloy and robust poly, with camera dimensions of 155mm x 152mm, a weight of 1520g and an excellent operating temperature range between -30 and 60C. Power consumption and current draw for 12V DC are 0.9A with a maximum of 10.8W, while PoE is 0.36A to 0.23A, with a maximum draw of 12.9W – these are modest numbers.

Test Driving Hikvision Ultra Dome

It’s a joy to power up a CCTV camera and – after resetting the camera, changing the server range, changing the camera range, re-changing the server range, calling up live view – discover it’s an absolute ball tearer. Something to note – camera images in this story are scaled by SEN website, but not smushed, so if you right click and save them, you’ll be able to see the quality as I view it at a file size of around 3MB – it’s worth doing this even with just with one image.

I set the Hikvision Ultra dome camera up in the usual complex street environment in Sydney. It’s the only camera on the 1Gb test subnet and is running through SEN‘s Dell Optiplex 9020 server. While I’m setting the camera up mid-afternoon, a crew arrives to work on power lines outside the office and there’s a lot of activity to a considerable depth of field, which makes testing easier and much more fun.

It’s about 3.30pm and the light is muted with shadows between the buildings. Colour rendition is excellent, with strong reds, greens and oranges. Depth of field is significant. The way the dome camera is positioned makes it a little tricky to see all the way to the end of the street as we might usually do, but given the slightly longer focal length at the wide end, the tilt angle chosen is best for this application. We are getting strong performance to 25 metres on both sides of the street and can almost get a look over our shoulder down the hill.

Looking at the monitor I can see face identification is court admissible to 15 metres and there’s strong recognition to 20 with considerable detail deeper in. I think the useful depth of field extends through 25 metres towards 40 metres, with worthwhile situational awareness deeper still. I have no trouble ascertaining what’s going on outside the office through my entire angle of view, with people moving around all over the place.

Along with the power crew and pedestrians, there’s a technician working on the house on opposite side of the street. He’s perhaps 20-25 metres from the lens but I have considerable detail of what’s happening. Looking at the monitor again, depth of field is highlighted with useful images of a group of people at least 50 metres from the lens – I’m able to see shirt colour, the style and colour of shorts, the colour of footwear, jacket colour, even skin tone.

A pedestrian walks across the scene at about 14 metres looking up at the workers replacing the power line. There’s considerable detail down to fine crinkles in clothing, and not just shoe colour, but shoe type. Carried items are easy to discern. I note the camera does very well in partly lit and shaded scenes – it’s creating a very level image when it comes to colour rendition, and the sharpness is good too.

Taking a look around the scene, there’s plenty of fine details around moving leaves – I don’t notice any of the smudginess you sometimes see. One thing I do notice is that the hyperfocal distance with this camera wound in ever so slightly is a touch longer than usual – maybe 1-1.5 metres.


There’s loads of detail across these images, including recognition of the workers leaning on the van 25 metres from the lens…

Some strong light appears on the far side of the street. The technician over there has got his ladder up against the second floor of the terrace and is preparing to climb, I can see his tool belt and make out certain items in it. The techs over there are doing some cabling work – maybe connecting power from the pole on my side of the street to this home.

Across the scene detail is exceptional, with face identification confirmed to 14 metres, near ID and definite recognition possible for another 10 metres, and plenty of useful detail of faces and clothing much further in. There’s no trouble whatever recognizing vehicles and vehicle details well past 25 metres. While this performance on the fringes is welcome, the performance in the surveillance sweet spot from 2-15 metres is the eye opener – it’s really good.

Something I notice as bright light comes and goes on the scene is that is has an impact on overall image colour temperature – the temperature is warmer when light hits the scene then goes back to a cooler temperature as light fades past a certain point. It’s quite a distinctive shift and I visit settings to make sure I am on automatic, which I am. With the cooler temperature I note more detail on the far side of the road.


In the raw image top I have court admissible face ID of the guy in the tie-dye hoodie crossing the street at least 22 metres from the lens. 

Next a vehicle goes by, and I have its plate at about 20 metres from the lens. It’s moving slowly, maybe 15 kilometres an hour as it moves around the parked truck, but it’s still good performance from a dome like this. Typically, you don’t have such a high level of activity throughout the scene – it’s great to be able to put the camera through its paces in this way. A Maserati goes by – this time I don’t have the black and white plate at 20-plus metres – unsurprisingly at this resolution and with this wide-ish lens focal length.

Performance from 15 to 20 metres is very strong. And that translates into strong performance deeper into the scene as well, looking at a worker at perhaps 30 metres perhaps more, I can see boot colour, apparel, worn items. If I could see the face of this person – which I can’t because of camera tilt – I could see hair colour and the presence of eyewear.

Now a woman walks past close to the camera at around perhaps 8 metres from the lens and I’m able to see some details of the logo on coffee cups that she’s carrying – details of her clothing and bag and hair are very sharp. I should point out that this is not typical performance for a dome camera and my streetscape just keeps on delivering grist for the Ultra dome’s mill. I’m able to see that in the right circumstances of light and subject approach angle this camera is giving me court admissible identification out to 20 metres. That’s exceptional for a fixed dome camera with a mildly wide angle of view.

I do notice digitally zooming in on this image that there’s some chromatic aberration around high contrast areas throughout the scene – it’s not significant – when I pixel peep, I decide it’s only 3 pixels wide. Another pedestrian walks past at 8 metres from the lens and I get excellent detail of clothing, carried items and shoes but what’s most relevant is that at the same time I have a face at least 20 metres from the lens that’s also court admissible identification.


Situational awareness in these scenes is excellent.

The same judgement applies nearly applies to workers on the far side of the road staring up at the power lines as a couple more slow-moving vehicles go through the scene. And I clarify that I have a slow-moving plate 20 metres from the lens – perhaps 22 metres from the lens. In the same image is a recognisable person walking down the hill at 22 metres carrying a bag and wearing white shoes and socks – obviously going to the gym – further in is a worker and there’s another vehicle whose plate I can see. The sharpness of the image is something to behold, and it’s reflected in the high levels of detail in foliage in the grain of the tarmac and in reflections from vehicle paintwork and windows.

A pedestrian is coming by at probably 7 metres from the lens. This image has got to be one of the sharpest and best balanced images when it comes to skin tone and colour rendition that I’ve seen in quite a while. It’s really very good. At the same time, I can see 22 metres from the lens a person is standing in white and blue runners wearing jeans and a grey sweater looking at a smart device. The person has dark hair and a mid-tone skin. On fast side of the street, one of the techs is carrying away a ladder while another talks on the phone.

It’s now nearly 4 o’clock and there’s a bright spot of light testing the camera’s ability to manage backlight variable light I either switched back to know the coolness and warmness is a measure of the clouds moving across the scene. The camera handles these variable well.

A woman walks by 8 metres walking brown dog wearing blue runners. The dog has a leather collar, leather leash. The woman is listening to her ear pods. She wears dark coloured glasses, a white top and blue jeans. Her face is court admissible. In the same image at about 22 metres a tall man in black footwear, black jeans, a black T shirt and dark glasses, with dark hair and a beard shadow is coming down the street. You can see his skin tone and ascertain that he has tattoos on his right arm. Beside him the person from earlier is pacing talking on their phone.


Easy plate in fading light…

It’s a reach to say I’ve got face recognition at this distance but if somebody knew this man, they would recognise him. In the same image on the other side of the road a worker leans against the wall with hands and pocket – a woman with brown hair and fair skin. In the house opposite the electricians continue with their cabling.

The next scene is similar and highlights again the excellent sharpness of this camera. A person walks by with a dog at 3 metres from the lens and the level of detail is exceedingly good. You always expect good detail within 8 metres of a lens but this detail is otherworldly in the full res image, compared to your standard affordable dome camera. In the same image at 22 metres a couple of workers are leaning on the front of a truck. A truck goes by and I note the absence of motion blur looking at the live stream – further I notice it’s actually possible to watch the power workers and comprehend their planning as they remove cable and replace it. The level of detail around their work site is exceptionally good – tool bags, tool belts, gloves, face recognition.

 


Focal length is 4mm and the camera’s image quality seems to compress depth of field – pedestrian is 6m from lens, worker on road is 40m. 

If it seems like I’m saying the same thing over and over, that’s because I am – this camera is so impressive that repetition feels like a process of discovery. In my busy street scene people just keep moving around making the Hikvision camera look good. Obviously, not every camera we’ve tested out here has had the benefit of this sort of activity. But even still, Ultra is right up there among the very best dome cameras we’ve looked at in this application.

At this point a man comes walking past at about 7 metres from the lens. At 14 metres from the lens another man comes walking down the street – I have face recognition of both these 2 that’s court admissible identification. But what’s interesting is that a couple of metres behind the second man is a third man and I have face identification of him, too.

At 4.15pm I back out fully on the zoom and pan the magic arm to increase my angle of view, as well as tweaking tilt so I can see further in. By minimizing the amount of street front on my own side of the street and I get more of a look down the hill to the right.

Sitting back down at my workstation I notice for the first time that I’m losing moving plates – a Porsche goes past I can see all the other details of the vehicle, but not the plate. Regardless, I still have good detail of workers over the road as light continues to fade. Just to make things more interesting, a fire alarm goes off in the building next door and fire trucks turn up this gives the opportunity to view a bit more activity and detail remains strong with little blur.

Low Light Performance

At 5.10pm a small truck goes past and I can see the image is beginning to break up a little – I have motion blur and tonemapping beginning to make their presence felt. However, there’s no issue with pedestrians in this light level and I have considerable detail of them. At 5.30pm a fast-moving vehicle goes by and looks like a Range Rover, but I don’t have sharp edges. Now it’s 5.40pm and beginning to get dark – the images are becoming a little flinty in tone. While I can’t discern plates I notice that I still have make and model of vehicles going by.

As we get towards 6pm it’s nearly dark. Interestingly, camera performance remains very strong, particularly in the presence of vehicle lights. Depth of field is good, too. I can see a person walking down the hill and in the background behind them somebody else is standing looking at their phone wearing grey trousers, a black jacket and black shoes. I couldn’t say I have court admissible identification at 6pm but I certainly have hair colour, hair style carried items, shoes, a tie.

At this point, I also manage to jag a slow-moving plate – admittedly the car is only going 5-10 kilometres an hour but still as light falls the scene gets a little 3-D with static areas rendered in very high detail while moving areas are softer. There’s a colour cast at the far end of the street where the sodium streetlights are but it’s pretty well balanced, considering.

As light continues to fade, the nature of the image alters with a little more noise pushing through, however, I retain considerable detail, particularly of pedestrians, and I’m able to assess car type, colour, model and my depth of field remains strong – it’s still useful to 25 metres and invaluable closer.

At 6.10pm a worker from a nearby building comes out of the tower and gets ready to ride his bicycle home. As he puts on his helmet, I get a snapshot that gives me face recognition at 15 metres in sub 2-lux colour. And he rides away, the camera does a pretty good job of keeping him together to the point I can see spokes in his wheels. Shortly after, a woman walks with past her face angled away from the light, yet I have court admissible face recognition again at 8 metres with sub 2 lux at the lens.

Depth of field remains a strength – there’s considerable sharpness a long way in – I can see street furniture, a lot of vehicle detail, foliage, a man in a dark gray suit with a tie on the far side of the road at probably 30 metres from the lens looks to have a fair skin tone and brown hair.

Conclusion

With its slightly long focal length at the wide end, decent resolution and great colour rendition, the Hikvision’s DS-2CD3747G2-LZSU Ultra dome really is a peach. In my busy street scene, it gives useful images from 2-50 metres with a particular strength between 2-20 metres.

The camera goes on impressing as light levels slide down under 10 lux, albeit with some motion blur with my slightly elevated 1/50s minimum shutter setting. You don’t get plates in low light unless they are angled directly towards the camera, but in those situations, perhaps with some help from the 40m warm/white light, Ultra Dome has the ability to surprise you.

This camera’s capabilities are such that it’s ideal for situations in which strong performance is required 2-40 metres from the lens across the angle of view. The camera is strong zoomed in but it’s also exceptional at wide settings. Nudged in just a touch it performs brilliantly on the street, delivering a 90-degree angle of view, face identification to 15 metres, face recognition to 22 metres and profound situational awareness to 40 metres. This camera is well worth a look.

Features of the Hikvision 4MP Ultra Dome include:

* Anti-illuminator-reflection layer
* High quality imaging with 4 MP resolution
* Efficient H.265+ compression technology
* Water and dust resistant (IP67) and vandal resistant (IK10)
* Clear imaging against strong back light due to 130dB WDR technology
* Audio and alarm interfaces available
* Focus on human and vehicle target classification based on deep learning.

 

#SEN #SENnews #security #electronics