Panasonic I-PRO WV-X1551LN AI Camera Review
Panasonic’s i-PRO WV-X1551LN AI bullet camera is a polished new offering from the Japanese manufacturer, with great build quality, solid physical and optical specifications, and an evolved AI engine that’s designed to enhance security applications in the real world.
The i-PRO WV-X1551LN AI bullet features 5MP resolution, a fast F1.3 varifocal lens with a range of 2.9–9mm, minimum illumination of .07 lux colour .04 lux in monochrome and 0 lux with IR activated. There’s also adaptive black stretch, back light compensation (BLC), fog compensation, highlight compensation (HLC), digital noise reduction, auto backfocus, integrated IR, image stabilistion and loads more.
But the heart of this camera is its AI capability, with 3 AI functions available, including the ability to detect suspicious changes in captured scenes. Depending on your settings, AI detects, identifies and determines movement, recognises the presence of human faces and captures them in the most visible state, locates moving objects, identifies where there’s no movement, and performs efficient data compression. Further, AI automatically adjusts image settings of the scene being analyzed, and optimizes video compression to conserve network bandwidth and server storage capacities.
These new i-PRO X-series cameras are available in outdoor and indoor vandal-resistant domes, as well as bullet configurations. Shared features include H.265 compression and smart coding to increase bandwidth efficiency and reduce storage requirements; colour night vision for use in low-light conditions; and Intelligent Auto (iA) with Super Dynamic functions for enhanced dynamic range in scenes with varying lighting conditions. All i-PRO X-series cameras are bundled with a Video Insight VMS camera license and come with a 5-year warranty.
Panasonic’s Matt Terrey drops the i-PRO bullet camera off at SEN’s office for testing and points out the new white colour scheme, which is a departure from Panasonic’s traditional silver livery. He also gives me a quick tour around the camera’s AI functionality in the AI-VMD setup menu. Watching him work, it’s immediately clear that Panasonic’s AI is a deep dive, offering plenty of nuanced functionality for operators in larger sites or on the street.
Panasonic I-PRO WV-X1551LN AI Features
Let’s take a look at the features of this camera before we get going on our test, looking at the physical specification first and the AI parameters after. This is an external day/night (ICR) bullet with a cast alloy housing that’s rated to IP66 against water and dust ingress and to IK10 and NEMA 4X against vandalism. The camera’s front panel has Panasonic’s ClearSight and Rainwash coatings, which push away water and the dust it’s often carrying to ensure a clear view. It’s got nice hand-feel and the new design is pleasing, too.
The image sensor is a 1/2.8-inch CMOS that has a minimum scene illumination in colour of 0.07 lux, 0.04 lux in black and white and 0 lux in monochrome with IR activated. Maximum resolution is 5MP at 30ips and the camera has a bunch of familiar features, including Super Dynamic WDR, adaptive black stretch, backlight compensation, fog compensation, highlight compensation, digital noise reduction, auto back focus, auto focus and focus assist. There’s image stabilization, corridor mode, 8 privacy zones, and iA Intelligent Auto functionalities.
The camera’s fast F1.3 varifocal lens has a focal range of 2.9-9 mm (34 to 106-degree angles of view), which is perfect for applications on the street or for complex environments, especially when the camera is elevated. DORI numbers for wide end detection are 25ppm, for observe they are 62.4ppm, for recognise they are 125ppm and for identify, they are 250ppm. Assisting with setup, the camera mechanism pans through 180 degrees and tilts through 100 degrees.
Compression includes H.265, H.264 and JPEG, with transmission in H.265/H.264, and other functions include constant bit rate, variable bit rate, adjustable frame rate, best effort and smart coding. There’s audio support via a built-in microphone, as well as an audio input and output, SD and SDXC memory card slots, external I/O terminals, an internal dehumidification device, operating temperature from -40 to 60C, power by 12V DC or PoE, the camera supports ONVIF S, G and T profiles and encryption of transmission is FIPS140-2 CAVP compliant. An SDK allows tweaks for custom applications.
On the video analytics side There are 3 AI functions available through the camera: detecting suspicious changes in captured scenes, optimizing the image settings of the camera (based on captured scene analysis) for better image usability, and optimizing video compression through captured scenes to save bandwidth at a throughput that’s 50 per cent lower than existing cameras while maintaining image quality.
Running the AI show are Panasonic’s AI-VMD, AI-Pryvacy Guard and Intelligent VMD (i-VMD). Something that’s cool about this camera is that AI-VMD lets you click to select detection objects you want highlighted on-screen – vehicles, humans, bicycles, etc. Importantly in some applications, AI Privacy Guard can pixelate faces and/or human figures to protect personal identities, while at the same time alerting operators to a human presence. According to Panasonic, the camera is able to deliver accurate VMD performance because it has an integrated AI processor, allowing it to identify a person, a vehicle or a bike and to send alerts to operators on the basis of those types.
Meanwhile, Panasonic’s AI-Video Motion Detection detects loitering, line-crossing and movement direction, while reducing false alarms caused by swaying trees and plants, shadows and reflections, and rain. And the camera’s AI Intelligent Auto optimizes image quality settings, including shutter speed, auto exposure, gamma correction, and noise reduction in real time to support these selections, while AI Smart Coding optimizes video compression to minimize bandwidth and storage requirements.
Test Driving I-PRO WV-X1551LN AI
We’re testing the camera on SEN’s venerable Optiplex 9020 server. The camera is powered via a NetGear 108SP 8-port PoE switch and is alone on the network. The camera is running at 2560 x 1440 pixels, has a maximum bitrate setting of 14Mbps (though we never approach it), and we are at H.264 to save the browser from frame drop. I should point out we tend to favour H.264 quality in our tests. Finally, VBR is activated and everything else is default.
My first feeling after hanging the camera out the front (and this is a compliment) is that the i-PRO renders like an Axis camera – the stream has that particular look about it – right down to the slightly more than usual barrel distortion and the latitudinal and longitudinal chromatic aberrations. Levels of detail are high, colour rendition is true and the camera deals with the bright parts of the scene 50+ metres from the lens extremely well. There’s no over exposure and no loss of detail in adjacent vehicles.
It’s a winter afternoon in Sydney, so what remains of the sun is away to the North and there’s shade on both sides of the street. The shutter speed has eased a bit, so I’m not getting moving plates at full wide. Depth of field is excellent thanks to that 5MP resolution, but the way variable light is being handled contributes to image quality, too. We often lose deeper detail in the afternoons thanks to over exposure but not with this i-PRO. I notice bitrate ranges from 5500-7000Mb in H.264, depending on movement across the scene. I’m running at ‘fine’, so bitrate could be reduced if required but there’s no pressure to do so.
My first move is to hop into settings and dial out most all the barrel distortion – we go from maybe 12 per cent to around 4 per cent. I check the other settings and leave them as they are. Going back to live view I notice that while plates aren’t attainable at this focal length, I have court admissible faces from about 12-14 metres.
I also notice the way the iVMD is handling AI designated targets in the scene. Cars and trucks are labelled vehicle, people are labelled people, with a red box around them and a green line showing direction of approach. Things need to be moving in the scene for a second before they are categorised by iVMD, though later on I hop back into settings and speed detection up of .2 of a second. If objects are static for a while, the box disappears, then the AI recaptures them once they move again. Detection range is surprisingly good – out to 70 metres at the widest lens settings today. Cool is that i-PRO is able to detect 2 people even when they are walking side by side at an acute angle.
Something I’m really interested in with this i-PRO camera’s VMD is just how capable it’s going to be as light levels fall. Obviously, getting court admissible faces and plates is wonderful if you’re seeking prosecutions for criminal behaviour. But it you are only monitoring general behaviour – perhaps keeping an eye on movement around a university 24 hours a day – then AI that can detect movement in scenes the human eye struggles to unravel is going to be valuable – and it’s going to cost less than thermal, too.
At 4.45pm, light is well and truly slipping away – it’s close to full dark by around 5.30pm in Sydney in winter. Colours have faded, there’s a little more blur around speeding vehicles but depth of field is still strong, and AI detection of humans and vehicles is as good as it was earlier in the day. At 4.54pm some one drives up the street in a white Audi A3 sedan with side skirts and a sunroof at about 80kmph (it’s a 40 zone). The camera shows blur, but there’s no question about the sort of car it is – nor what sort of bloke is driving it. As evening turns to full night, I notice that cars are less often identified but humans almost always are – later I see this comes down to my settings.
Next day, I drive the camera in the presence of 65,000 lux of sunlight, which I zoom into using a focal length of around 6mm. This is a good focal length for the camera and for my street application, giving tons of detail, brilliant depth of field and retaining strong handling across a scene combining shadow and strong variable light. Something I notice at this focal length is those widespread chromatic aberrations ranging from 3-5 pixels deep in areas of highest contrast. In these conditions there’s rapid detection and recognition of objects moving in the scene.
AI detection range is excellent…
Throughout the day, I appreciate the colour rendition, the depth of field and the extra detail afforded by the longer focal length, which has tuned out most the static parts of the scene I was covering at full wide. At around 3.15pm a person comes running down the street and is tracked by iVMD, while the camera delivers court admissible face recognition. By 3.30pm on this second day of testing I notice that I am able to get slow moving license plates, as well as recognise faces behind windscreens at around 20 metres from the lens.
Attracted by this, I zoom in further still – I’ve lost the pavement on my side of the street and am covering a section of road between 15-25 metres from the lens, with a shortened view of the pavement on the other side. With the aperture stopped down a little more I’m not able to enhance moving plates but faces at this focal length are exceptionally well rendered, with accurate skin tones and considerable detail of attire and carried items. Motorised bullets are good like that.
Zoom concentrates pixels in a very pleasing way…
At 4.32pm I’m surprised to find the camera’s AI is able to track pedestrians even when they are behind trees and branches at a distance of around 30 metres from the lens. By 5pm faces have been lost and I have motion blur and some tone mapping around moving objects – the latter more prevalent around the bonnets of fast-moving cars or people’s legs. Even so, I retain high levels of detail of vehicles and pedestrians, and the AI goes on putting objects into boxes as light falls completely away. Something I appreciate is that this detection and categorisation also affords a sense of context – it allows you to see how far people are away from each other.
I switch to monochrome with IR activated next – operational IR range is at least 40 metres, with a nice spread. With IR activated, the shutter drops further still, and I lose detail of faster moving objects, like motorcycles, though people and slow-moving cars are still to considerable extent, identifiable.
Something else that’s identifiable to the camera are people I can barely see. It’s fair to say the camera’s AI engine doesn’t always get a person instantly but as they move through the scene it will find them and in some cases, it tracks pedestrians from under the lens out to about 40 metres away in sub-4 lux, which I wasn’t expecting. It’s fine performance from the i-PRO. Finally, I go back to colour and notice more detail, including better face recognition. Under the lens – around 5-6 metres away – it’s near court admissible with subjects approaching the camera point. If the camera were to live in my application, colour is where settings would remain.
IR top, unassisted monochrome at the bottom…
Panasonic’s i-PRO isn’t only about camera performance. You hop into AI-VMD via the external software tab in the tree directory. Once you’re inside, you start out by selecting a detection area/programme you want to monitor by delineating a colour, then click vehicle, human or bicycle icons to tell the AI engine what to look for in that area. There are 8 possible detection areas per camera view in each of 2 detection programmes.
Next, you hop into Detailed Settings, where you can tweak motion sensitivity, as well as the sensitivity of objects being detected. In my application, the settings are slightly lower than standard for motion, about 80 per cent higher for humans and about 10 per cent lower than standard for vehicles, with bicycles around standard settings on the slider.
What’s interesting when you start playing with these settings, is that the i-PRO automatically adjusts camera settings in support, including things like Super Dynamic, white balance, DNR and auto contrast. Next, you can schedule settings – these include detection zones for periods of operation – my 2 zones are set for 7 days, 24 hours. Finally, there’s alarm settings – here you can select email notifications, FTP transmission, recording settings and alarm settings.
Panasonic’s WV-X1551LN 5MP outdoor bullet is well-made and offers solid optical performance across a range of conditions. In lower light levels, the best performance is to be found in colour down to around 4 lux. The camera offers strong colour rendition, excellent depth of field and has a comparatively fast zoom and re-focus function. The big resolution delivers more detail than we are used to seeing and it comes with little discernible impact on low light performance.
The camera’s overall performance is considerably enhanced by 3 very polished AI systems, which work together to deliver real operational benefits, using motion detection, object identification and optimisation of camera performance. Best of all, AI-VMD is at once capable and relatively simple to apply and its functionalities make Panasonic’s new i-PRO range worth careful consideration.