Zones and triggers not working consistently

[elbillete]

Hello,

I have been using BI for a few years now, recently upgraded to V5. I record 24/7 on all my cameras so went with default config and no zones. I am now trying to setup LPR and decided to go with zones. I am also not recording 24/7 on these but rather when triggered. In trying to setup zones and crossing, it seems I am not getting consistent alerts. When I say alerts I mean a video clip under alerts, not a notification of any kind. I am comparing the LPR cameras with the alerts I get on the overview cameras. I have tried different setups and still miss some of the traffic going down the street. I setup the cameras as follows to see what config would give the best results:

I setup 2 cameras with zone A covering the street only. No zone crossing configured. I get most, but not all traffic going in either direction. I have two cameras configured this way. One facing North and the other facing South. They are about 100 ft. apart and cover the same street in front of my house. They both behave the same as far as inconsistency in the alerts. I may see an alert on one and not the other.

Setup clone 1 of one of the cameras above with half the street on zone A, the other half of the street on zone B, and the entire area, including zone A and B in zone G. Also left a small gap between zone A and zone B. Crossing A>B. I understand this is the correct way to do this, however I get practically nothing from that camera. Every now and then I get a stray walking their dog. Literally I may get one alert a day.

Setup clone 2 of the same camera. Similar with half the street on zone A and the other half on zone B. This time I left no gap. They are touching or even overlapping ever so slightly. Again crossing A>B. This setup get most of the activity moving A>B but again not all of it.

All of the above information is for daytime. At night all bets are off! I may get an alert once maybe twice a night between 7PM and 7AM but most events do not trigger an alert. There may be a lighting issue, but that is something I can address as I move forward. My concern now is getting it to work in the daytime before I address nighttime.

I will add that the LPR cameras are Dahua vs the overview that are Hikvision. I am not necessarily comparing one to the other simply using the alerts from the Hikvision to match and see what is missing from the Dahua.

What if anything am I neglecting? Why are the alerts inconsistent and missing? Is this even a BI issue? I have seen the cameras capture people walking by to large trucks and everything in-between. Yet it misses a vehicle every so often with no pattern that I can come up with. I have gone over the help files and all the videos I could find but still not getting the results I need. I am not in favor of one zone config vs. another just one that works. Can anyone provide any clarity as to what may be going on? Any advise or recommendation is appreciated.

 

[fenderman]

change the algorithm to simple

 

[elbillete]

change the algorithm to simple

Thank you for the quick response. That seems simple enough. Can you shed any light on this? I changed one of the cams to try it, but wouldn't mind knowing the why. I read up on the help file but not much info except that it emulates V4.

 

[DsineR]

Edge Vector is a relatively new algorithm, works great with a single zone but not so much with zone crossing.
A few threads on this topic here & here.

 

[elbillete]

Edge Vector is a relatively new algorithm, works great with a single zone but not so much with zone crossing.
A few threads on this topic here & here.

Thanks for the link. That clarifies things a bit.

I went ahead and changed the two clones to simple algorithm in the hopes that it would resolve the issue. It may be better but it is still inconsistent compared to the overviews and to each other. I don't understand how the same camera with an almost duplicate config can give different results. The images below are from the setup in clone 1 as described in the original post. Clone two is just the road with no gap between A and B and no zone G. I am getting about equal results from both clones since there is the occasional trigger that is not captured by clone 2 that is picked up by clone 1 and vice versa. I have made a lot of changes to the clones and maybe the sample size is quite small. I will give it a few more days and compare results.

I have never taken the time to compare results between cameras. I have always recorded 24/7 and never worried about it. Now that I am only recording when triggered I see different results, even when comparing the overview cameras. Is this normal? Shouldn't all cameras trigger on the same event? What am I missing?



View attachment 76069
View attachment 76070
View attachment 76072

 

[elbillete]

This is really just a zone crossing not working as expected. Can anyone provide any input?

 

[wittaj]

I have experienced the same thing with a cloned camera and not changing a single thing. I chalk it up to the computer or camera can't keep up and results in a tiny lag or it could be BI or some combination. Maybe it processes the first camera and then the clone because the vehicle will always be in just a slightly different location in the snapshot between the camera and clone. And occasionally I also get where one gets the trigger event and the other camera didn't trigger for the same event. Now mine is for a plate reader that is zoomed in tight, so a fraction delay and a fast moving car and it can miss it. At least that is my experience.

 

[elbillete]

That's exactly where the issue lies. These are also LPR cams. If we miss a few frames, we miss the entire event. My idea of the clones was just to confirm which zone crossing worked better, not that I needed different functionality from the clones. I have no issue deleting all the clones and keeping just the cam. However, at this point I have little confidence that either setup will capture all events. Like you I see missed events on both clones and on the cam but it is more than just occasionally. I miss at least 1-2 on each clone daily. To me, that is alot.

 

[wittaj]

Yeah, I found mine out when I was testing different plate recognizer programs and wanted a clone camera just so I could just delete that whole camera and files after I was done testing and that is when I noticed the issue. For all other applications it probably isn't a big deal.

I found this to be the setup that worked best for me as the vehicles were not in the frame long enough in two different zones for it to reliably recognize it. The one zone approach at this location allowed me to catch the front plate of the cars going right to left and the back plate of the cars going left to right. Occasionally I would get a large truck or a very fast vehicle that was missed, but I also run this continuous motion in the event that happens or it missed the motion.

 

[elbillete]

That's a good idea. I did not consider taking a somewhat minimalist approach. Interesting that the more complicated form of zone crossing with 2 zones, a gap, and an overall coverage zone is the one that captured the least amount of traffic. Maybe this is the solution. KISS principle at its finest.

I wish I could go continuous instead of when triggered. My CPU sometimes goes ballistic, so probably not a good idea until I figure out what is going on.