********** *** ********.
** ******* ****** ********** **** ******, ignoring **** ****** *** ****** ********** components ** * ******, *** *** full *********** ** **** ******. ******, it *** ****** ** ***, ** have *********** **** ** *** **********.
Classic ********** ******
**** ****** *******, ********** *** ****** defined ** **** ****** (*** ** line ****** ** ********** ** *****), which ******** *** ****** ******* **** the **** ****** ***** *******.
** ********, **** ****** **** *** from *******. **** **** ********** ***** lighting ********** *** ******* ******* *********. Plus, ************* **** ***** **** ****, leading ** ****** ******** ** *********, but *********, ** **** ****** ** 600 ** *** ** *** ***** ** ****.
*******, ** ***** **** ****** ********* to ******* **** ***** ***********, *** just ***** ***** ** * **** (i.e., ****** ** ****** = **********).
The ** ** **** ****
**** ** ******** **** *** ** line **** *** *********** ** ** IP / ** *******.
*** ***** ** ** *** ***** chart **** * ****** ********** ** capture *** **** ** *** *****. The ***** ***** ***** ** ** camera ***** ******:
** **** ******** *** ****** ** TV ***** ********* ******* ********* ** various **********. **** ** ** ******* of *** **** ******:
****** **** ** ***** *, *** number ** ** ***** ******* ** ~450-500 ***** ***** * ** *** right ** **** ******, *,***-*,***.
Questions ********
**** **** ******* *** ********* *** questions:
Key ********
**** *** *** *** ******** **** this ****:
******* ****** **** **** ** **** included ****** ** *** ** **** report, ** ******* ** ***** ************, since ************** ** ***** ******* *** be ********** *** **** ******** **** person ** ******.
Test *****
**** ***** ***** ******* *** ******** setup ** *** **** *** *** of *** *****.
Summary ********
**** ***** ***** *** *** ** found *** **** ****** ** ********* lighting **********:
***** ** *** ***** *******, ***** count ******* ** *** ** * ratio ** ************* *:*.** ** **** light. ** *** *******, ***** * 5MP ****** **** * ********** ********** of ****, ** *** ******** *** captured ** ** ~*,*** (******.**=*,***).
Pixel ***** **********
** * ********, ** *** ** our ******* ** * ******** *** room ******* ******** ********. ******* **** tested *** ** * **** ** properly ****** *** *** ******** ***** to *** *****, *** ******* ******** were ****.
** **** ** *** ********** *****, TVL ****** **** ~***-*** ** **** to ~*,***-*,*** ** *** **** ******.
Lowering *****
****, ** ******* *** ****** ** 3 *** ** **** ********* *** captured ** ***** ****. ******* **** standardized ** */*** ******** ** ***** low ***** ******.
*** ******* ************* ** *** *******. The ***** ****** **** ********** *********** in **** ***** *** ** * combination ** *** ***** *********** *** resolution.
(***** *** *** **** **** *******)
**** *** ****** ***, *** ***** level ***** ** ***** *.* ***. Only *** **** *** ***** ******* produce ****** ****** ** **** *****, again ** * ******* *********** *****, with *** **** ****** ******** ** ~200-300 ***, *** ***** ****** ** ~550-600.
***********
** ***** ** **** ***********'* ******* on *** ********, ** ****** ******** from ******* ** ******* (*-***, **** a ******* ** ** ** *** Axis ****** **** *****). ***** ******* had ****** ****** ** ***. **** compression ********* **** **** ** *** edges ** **** ****** ******* ** the *****, *** *** **** ***** used ** ******* *** ******** ****** unchanged ********** ** ***********.
*******, ** ******* *** *********** ******* H.264 *** *****, ***** **** *** no ********* ******, ******* ***** **** *.*** ** ********** *** ***** ******* than *****.
Lens ****
** ***** ** *** **** ****** lens ********* *** ** *** ********, we *** *** **** ***** * 720p ****** **** **** ********* ******: one ******** **********, *** ********* *****, and *** "***** **** **********" *****-********* rated ****. ** ***** **** **** selection *** *** ****** ******* ****** on *** ********, ***** ***** *****, with * ******** ** ***+ *** possible ****** **** ******** ********** ****** to *****-*********. ************, **** *** *** megapixel ***** ****** (******** *** *******) resulted ** ********* *** ******* (***-*** vs. ***-***).
****** **** **** (*-*):
Image ********
*******, ** ******* *** ****** ** image *********** **** ** ********, *********, and ********** ** ***.
***** *****, **** ********* ************* ******, ******** ******** *** **** it ** ******* ** ******* **** default. ********** **** ******* ** ******* provided ** **** ** ***.
******** *** **********, ** *** ***** hand, *** ** ********** ****** ********** of *******.
Members,
What should we do with testing tv line counts? Should we adopt this as standard practice?
Or should we adopt eye chart testing?
Or neither?
It is good to have a numerical metric to compare performance but it is important that the metric be meaningful and accurate.
I think this is useful and interesting data. As you point out, there is much more to getting detailed images than just simple pixel count.
For the cameras you test, it would be helpful if you had a summary of a few key components in a standard format:
Max TVL Measurement (though I wish the TVL term would die)
Max sustainable framerate at full res
Color accuracy
Power consumption
Key protocol supported.
WDR range
Other people might have additional suggestions for data, but the idea would be to summarize all tests with a set of consistent measurements that allows people to make semi-informed comparisons between various units.
Both have good uses in tests. This test reminds me of how dpreview does their camera and lens tests. These tests may be better on long lens distortion perhaps.
By far, I found the TVL comparison and testing more beneficial. I could image a benchmark being established, such as a Pixel count/TVL ratio.
If an product under test performed dramatically different than the de facto standard, that seems to me to be very relevant information. In other words does the camera ultimately deliver on it’s pixel rating?
John, have you considered the Rotakin test chart? It was designed specifically for CCTV performance testing. However I am not sure whether they have updates it to handle megapixel cameras or not.
Jason, we have considered the Rotakin but I have never understood its value. It's ~$1000 for a cardboard cutout and, outside the UK, it has no standing.
For quantitative measurement, I am not sure why that would be better than an ISO 12233 or a Snelling eye chart for that matter.
I would agree that it seems a bit pricey, but it does add some further dimensions to a camera test - movement, the motor is capable of spinning the figure continuously which may be an added criteria for your testing (motion blur and frame integration impact) - and the number plate characters may be UK standard but are very relevant for a lot of users who want to "read number plates".
Of course you could always design your own combined chart...
It's not the price to us, but the value. You can get real used license plates for $20 online. Plus, there's lot of others ways to check for blur, most realistically simple using a real subject and vehicle.
Interesting, in that it would appear that for most purposes 1080p is the optimal bang for the buck, performance wise. Nearly double that of 720p and only slightly less than 5mpxl. Also has the best low light performance.
Great Job Ethan
What Cameras were used in the test with which len's ?
Ok to respond privately . send me an email , John has my email.
We test cameras this way and use a test chart that Irlabs.net hands out. It has resolution on one side and colors on the other. We have been suprised how many subtle color varitions the different cameras produce. Many cannot tell the difference between dark purple and black. The reds also show a great variety of results. Your comments about the lenses differences are important also. Better glass, better picture.
As a comparison, I would like to see a higher end analog camera that is spec'd at 650 TVL in your test. According to what you show here, a 720P camera is worse than one of the analog cameras rated at 650 TVL (assuming thay they really have that spec.)
Hi Peter,
Thanks. We will put this on the list of things to test. I've never heard anyone say that 700TVL or 800TVL made much of a practical difference so I am not bullish on the results but worth a look.
Certainly I found this testing method beneficial but for my needs, I find your subject/eye-chart/field- enviornment method of testing to be of much greater value. Thanks to your team btw.
This software is freeware provided free of charge by the Olympus Corporation to which the creator of the software belong as a service to allow users to make more effective use of the ISO standard ISO12233
:)
I am an IPVM eye chart fan but only when I can actually see your test results. Life would be much simpler if if you gave us a number.....whether you call it TVL or something else. I think if the TVL method got me narrowed down to 2 or 3 cameras then I could go look at the eye chart stills to finish the decision making process, it would be perfect. I know that some of us want everything. Thanks for all you guys do!
A single number would be great but:
Overall, trust me, I see the value in a single number because there is real demand for it and it would make us look more 'scientific' (because people wrongly assume a precise number means a precise method). But I am concerned about the downsides.
Is there a way to introduce the effect of motion on your test? You mentioned there was no significant difference using various compression settings, but I would assume that was looking at a static scene. Start moving the chart in front of the camera, and let's see what happens. I assume that is where a 730p 60 IPS cameras will perform much better than a 10 MP camera, etc.
Greg, we have done various tests with motion, like 60fps vs 30fps one and the H.264 I vs P frame.
I am not sure what you are looking for here. Are you interested in seeing if there is going to be compression artifacts or?
For instance, in a test of 720p/60 vs a 10MP camera with motion, I am quite certain that neither will have any blurring or smearing but the 10MP cameras (typically maxed out at ~6fps) will miss very fast motion (like counting money).
Hi
This test is nice, but it's a B/W only testing,so we need additionnal stuff again.
The international EN50132 target is going to be an international test reference
It does combine B/W , Color and details ( Pixel density Inspection / Identification / Recognition / Detection ) and it's very cheap A3 format plastified sold here round 3 for 60$
Marc,
Thanks. Were on the chart is the TV Line count / pixel count / resolution portion? Is it the D / C / B / A portion and if so, what ranges does it cover?
In fact with the European Norm EN 50132-7 includs 76 pages explanations, and also the 9 UK faces ... and how we can test Licence plate reco ..so I include just small parts (in French sorry)
We are working inFrance on the the futur way to measure performance of installations and also maintenance process. Because the technical level is quite low (in general) we try to push the most simple way to test
Bellow kind of classification .
In France 90x60 pixels are required to reco a face (25x15) so +/- 400 PPM
"Compression
In order to test compression's effects on TVL captured, we varied settings from minimum to maximum (0-100, with a default of 30 in the Axis camera seen below). These changes had little effect on TVL. "
It seems you have changed compression with enough light.
Try to change compression on low light or reduce contrast of the test chart image then increase compression. I suppose in these conditions compression will have big influence on resolution.
TVL testing is what led me to IPVM. I see that this is from more than 3 years ago but not seeing any recent. Any plans to actually do TVL testing?
