Subscriber Discussion

Gige Vision Cameras - Industrial

Does anyone have any experience with these and any software they interface to? Had a customer ask me about it today and all I could do was a quick Google search on the basics of it. Just looking for some real world experience rather then Sales mumbo-jumbo online.

How does he want / plan to use it? He wants to record / manage them as part of a regular IP camera deployment or?

Related - Machine Vision Camera Options

That's what I'm trying to figure out. Being that is is uncompressed video I can't imagine any long term recording at all unless you have an unlimitted storage budget and your own data center.

Was just wondering if anyone had any experience spec'ing them as to what they are typically used for other than viewing live video of whatever the camera is pointed at.

On that note, here's a discussion about recording uncompressed video from industrial / machine cameras. Short answer - not really feasible.

I don't work in the surveillance field. We utilize uncompressed video in real time and utilize a ring buffer for short-term storage, but encode it to H.264 for long term storage. A few years back, we spent just under $10K for each four-channel H.264 encoder. Given the cost, complexity, and reliability challenges from this approach, this would probably be undesirable unless your customer feels there is some kind of compelling need.

Horace, what's the benefit in your application of using uncompressed video for real time monitoring vs 'regular' compressed video?

That sounds like a question for Todd. It is a key feature of HD-SDI/HDcctv as espoused by the HDcctv Alliance.

Some image processing algorithms and transforms have noticeable performance degradation when comparing their use on lossy compressed video vs uncompressed video.

Which image processing algorithms specifically?

Even something as straightforward as the Radon transform of a noisy image demonstrates degradation. The effect is most pronounced in those cases in which outcomes are noise-limited.

This can be seen easily by running a Radon transform on a test image that is noisy. One suggested image might be your license plate images which are corrupted by glare. When edges are crisp, performance with or without compression is likely to be adequate, but as clarity recedes, compare results of the same transform on an uncompressed vs compressed image.

In actuality, that was a (unfortunately unattainable) goal of mine in my initial research into replacing our monitoring and recording system. For as long as we have been deploying cameras, "live" images were superior to the recorded images. That was the case when we used VCRs and it continued to be the case when we tested and later deployed digital recording systems in 2002-2003. Our entire operation was based on that capability: provide the best possible live images while providing acceptable, but not ideal, images during playback. In good part, that concept was aimed at minimizing operator fatigue and maximizing productivity.

The decision to eliminate our matrix and analog monitoring was not easy. I knew we would have to accept lower quality live images of encoded analog cameras to allow the higher resolution images provided by megapixel cameras. No technology was capable of providing analog's straight-through video quality without requiring a huge expenditure on infrastructure and replacement of the vast majority of our cameras (over 1,000).

HD-SDI and HDcctv, while supposedly having the capability to utilize our existing infrastructure, never demonstrated the ability to be incorporated into a suitable matrix and VMS or to be simply transported over long cable runs and a mix of coax and single-pair UTP. HD-SDI matrices are a throwback to ancient technologies like Pelco's Compuswitch. 960H demonstrated no better resolution than a typical 600TVL traditional analog camera.

One intriguing concept that was seriously considered was Sony's SLOC hybrid cameras. The capability of viewing SLOC cameras live via our analog matrix while recording them in HD presented a concept that I felt was revolutionary: better playback images than "live". Unfortunately, SLOC really hasn't taken off within the industry.

In the end, we wound up insisting that all of our encoded analog cameras be streamed at higher bitrates than most manufacturers and Integrators recommended. While not the ideal solution, it does alleviate at least some of the problems introduced by encoding the analog cameras. Obviously, replacing a substantial portion of our analog cameras is a future goal but that will happen over a number of years.

Thanks for the information everyone, it is very helpful!