Please help me on my first project!!!

[alexis 0]

I was wondering if any of you could be of assistance.

 

I am just about set to start my first installation. I will be installing an 8 camera standalone DVR system. I was going to install:

 

3 Indoor Nuvico NVCC-SDD Dome cameras w/ 4-9mm varifocal

5 Outdoor Sanyo VCC-4794 Day/Night Cameras w/ 3.5-8mm varifocal

 

But...finalizing the project deal hinges upon one important criteria that the client has. The client wants two specific things:

 

1. Excellent resolution and video quality at night.

2. the ability to capture video at night that is high enough quality so that facial features can be reconized when the image is zoomed on.

 

the property has very little ambient light, with only a dim street light at the end of the block. The cameras will have to view objects no more than 25-30ft away maximum. Cameras wil be positioned at approx 13ft high.

 

I would greatly appreciate if some of you could provide me with two things

 

1. Can you recommend outdoor cameras that you have used that are either day/night or integrated IR that you know are excellent quality. Price range can be from 350-700 dollars.

 

2. Can you please provide any night time jpeg, image files, or video clips that you have of the cameras you suggest that you can post or email me.

 

I would love to start this project and hopefully share my experiences with others on this forum. Thanks!

[Thomas 0]

For number 2....what is the distance he wants to do it at?

[alexis 0]

If a person is captured on video up to 15-20ft away he wants to be certain that he will be able to zoom in on the captured image and have enough clarity to recognize facial features.

[Thomas 0]

That can be tricky and an exact answer is tough without seeing the site but some numbers to ponder.

 

If the lens is at max zoom (9 mm) and focused at 15 ft gives you a field of view 8 ft by 6 ft. That gives a 2D volume of 48 sq ft. The human head varies but we'll assume that it fills a volume of 1 ft by 1 ft or 1 sq ft. That works out to 2% of the image. Now we get to the fuzzy part. So with that 2% of the image...if my boss walks by, I can ID him. (I'm using more clues then just face; i'm using body type, mannerisms, etc.) If it's someone I don't know...then I lose some of the guess factor.

 

But if your question is: Can I use CSI style digital zoom to make that 2% useful? The answer is no. No magic tech in the world makes digital zoom worth a damn at the moment.

 

And how does light play into this: So depending on day/night or IR, you have to take shadows into account. What other light sources are there? Will the light be behind the person? Will the light hit the left or right side of the face or will it be frontal.

 

I don't mean this to be a smart ass answer, just pointing out that it's not a yes/no question.

[Cooperman 0]

alexis,

 

Everything Thomas says makes absolute sense, and that's without allowing for the fact that with cameras working 13' in the air, and almost certainly with the 'Gain' working overtime at night, I doubt that you'd be able to recognise anyone on screen at 25 - 30' distance.

 

You'll need to have a think about ideal camera positions, alternative lenses, and possible supplementary lighting if you're going to have any chance at all of producing good quality useable images.

[securitymonster 0]

geez thomas, you should post more often, I think we can all learn alot from you!

 

What about using a megapixel camera? I haven't messed with those so i can't speak but with something worth a darn and a good low lux cam, I would think you can take the image into a video software program and have a little more success........

 

Am I way off?

[cctv_down_under 0]

Most Megapixel cameras use CMOS chips which up until now are much less light sensative, agreed the resolution might help but Thomas is right and this is a common mistake indeed, the head and body must take up a large percentage of the screen..light will not be your major issue if you are propared to spend good money on cameras and lighting but distance most certainly will be...here is a tip I use to try to explain what thomas said!

 

Take a digital camera to site with a zoom lens, take a shot zoomed in and one zoomed out, make him hold something in his hand and place something a few feet in front of him...show the person that with the camera zoomed out you can see he has something in his hand and you can see more of the scene, but with it zoomed in you can identify but not cover a very large area, so you will need lots more cameras if he wants facial ID.

 

You can use binoculars to get the same effect, tell him that if you were at a horse race you could zoom in on te jockey and see his face, but then you would have no idea what was going on around him...or you can see the entire race and not identify the jockey!

[alexis 0]

Thanks for the information. I understand what you mean by distance betwen subject and camera and field of view. Taking into consideration that there is very little light(ambient, artificial, or natural) which outdoor box color cameras would you recommend for the price range of 400-700 that would provide the best image quality at night? Both day/night or IR.

Thanks.

[Thomas 0]

I don't post more because most of my understand of CCTV is theory based rather then practical experience. I have never once set up a camera in the field. I have set them up in testing environments but that's different. When it comes down to it, people like Cooperman know alot more then I do.

 

When it comes to Mega pixel cameras, they can do some of this, but you still have restrictions. I've used Covi, so I can pick on them.

 

I set up the camera and I get a 640 x 480 signal from them. And then I get two more signals at 640 x 480 that are zoomed in. But I still need the zoomed in areas to be in the right position for that to be useful.

 

The limits of any kind of enhancement comes from what ever the lowest resolution is. It's why a monitor always makes a camera look better then a time lapse. The monitors resolution is almost always greater then 240 lines that the tape uses.

 

This limit is why I can't take the first signal from the megapixel camera and do magic with it. Becuase my stream is 640 x 480, I can't make information appear that simply isn't there.

 

Now you can do some limited forensics work with cleaning up an image. But I'm not going to do much to add detail. I won't magicly see that the thief has a beauty mark if I didn't see that before.

[Cooperman 0]

This thread is actually throwing up some very interesting points.

 

In no particular order, cctvgeeknz mentioned that megapixel CMOS sensors are less sensitive, but the suggestion is often that it's due to CMOS fabrication methods rather than for example CCD construction.

 

My understanding on the theory of imager construction (which is most definitely not my specialist subject), is that the lack of sensitivity is a physical limitation relating to the significant reduction in pixel size - i.e. to cram more pixels onto a chip, you have to make them smaller, so they individually cannot gather as much light. There is no doubt that performance will improve over the next few years, but at the moment, high end CCD's do perform better under lowlight conditions, and that's probably what alexis will need for his particular application.

 

The analogy of watching a jockey is also quite interesting, because it demonstrates a very simple principle for identifying the ideal position for locating a camera.

 

If you are watching the horses coming around a bend with the jockey facing straight towards you, it's relatively easy to keep the face 'in camera' using a narrow telephoto type lens. If however you were trying to watch the jockey racing past from left to right (at ninety degrees to the direction of viewing), you would have to use a wide angle lens, just to have sufficient images available to see what was going on. A clear example of how important it is to establish the optimum location for any CCTV camera.

 

Thomas, you mentioned about theory and practice. There are people out there working as exo biologists who are highly unlikely to ever meet a little green man from another planet, but they are still experts in their field.

 

You consistantly demonstrate a degree of actual expertise in your chosen field, which most of us could only dream of.

 

One point which you touched on, which is rarely if ever mentioned relates to image enhancement. I have it on good authority that one of the biggest problems for image enhancement experts, is trying to recover detail from a recording, when that detail has already been compressed into oblivion by the DVR / NVR / Server. Although smaller file sizes such as those derived from MPEG-4 devices are an advantage in terms of storage economies, I suppose the old adage about there being no such thing as a free lunch, immediately spring to mind.

 

alexis, apologies for the waffle; coming back to your problem, have a proper think about where is the best place to locate the cameras, using the narrowist angle lenses you can get away with.

 

Then consider how you are going to light the area. Without knowing where you are, and what if any limitations there are on using visible lighting, I would have suggested you use (within your budget) a decent 'Day / Night' high res. box type camera, and seperate strategically placed illuminators to get the desired degree of coverage.

 

If there were any way you could post either some images, or a half decent plan of the area, then I'm sure we could come up with some more practical suggestions.

[Thomas 0]

It's an issue of smaller pixel size. CCD based digital cameras have the same issue. You can get around it by playing with lens and a few other tricks but you have accept larger camera sizes.

 

As storage takes off, I suspect we'll move away from MPEG based recording all together and move back to MJPEG and MJPEG2000 simply because you can clean those images up alot more. The knocks against them simply are that you have to accept lossy images to get decent file size. But when I have 8 terabytes of storage that costs $100 a terabyte and gives me 60 days...it doesn't matter as much.

[alexis 0]

Wow! Thank you all for the great information. You guys are very knowledgable and I hope to get to your level one day soon. I think that I will definitely consider using some combination of illuminators for the project as it seems that is the only way to ensure visibilty. I will most certainly try to post an image of the facility layout asap.

[replanet 0]

In my warehouse I have 4 sanyo 4594 that are less resolution than the 4794. I used the a fuji varifocal and my results have been great. I also use 3 sanyo 3972 not day night cameras and 2 samsumg sdc 311. My warehouse is around 120 ft long by 50 ft wide. I can tell you that achieving facial recognition of a strager is very difficult. In order to achieve facial recognition I had to focus in an extremelly small area, the door. I have most cameras setup for general view and then one only on the door for facial recognition.

For the outside of the warehous facial recognition was not achievable with only 5 cameras. I set four on general view and one for "car recognition" for my needs this was the best compromise.

I did a sketch and used the graphic cctv lens calculator and still had to move some cameras 2 or 3 times to get proper surveillance positions.

 

You professional do have a tough job, camera selection, camera placement, unrealistic customer expectations within budget. Thanks for sharring all your experience and experties with some newbies like me.

[cctv_down_under 0]

You consistantly demonstrate a degree of actual expertise in your chosen field, which most of us could only dream of.

Would have to agree with that comment.

 

I think the most important thing is that you are realistic in telling the customer, i often find myself asking if they want overview or identification cameras, and when they say "what is the difference?" I HONESTLY tell them it will be at least 3x the amount of cameras, they then want to understand and I talk them into throttle points.

 

To give you an idea, an open petrol station that wanted licence plate recognition, i zzomed in my digital cam and took a shot of a car driving in, i made sure to get a bit either side of it, we then went and measured it, it was about one 16th of the actual possible driveway entry so i told him you would need 16 cameras, he said tha would be expensive but i told him to cement off a border, so they had to come in through a channel (narrow driveway) this allowed one camera only and allowed me to be particular about light smear etc, the cost of the concreting was cheaper than the cameras.

 

The same applied to the bowser area, but I used a PTZ on presets to get id shots of people using pumps and a general overview camera as well.

 

the same goes for pubs, have one camera as id on the entry and then the others can be overview.

 

 

As for your car park, remember dark areas of solid objects (asphalt etc) have little light reflectance so although light may be at a certain level at a scene it needs to bounce back to the camera and if the reflectance is low then the light gatehred back is small, this is common for car parks, if it were me I would go for a Bosch Day Night cam with IR cut filter and a Computar low FSTOP IR corrected lens, If needed afterwoods I would then add IR illuminators....

 

 

Seriously just take a digi cam and a LUX meter (available at any camera shop) and show him the light level and the different scene views...they then understand