Dahua IVS questions

[marklyn]

I'm fairly new to using IVS. In the past I've always relied on Blue Iris to take care of motion detection and I've gotten pretty familiar with BI in that respect. Still, I'm liking Dahua's IVS on my 3 cameras and want to expand my use and knowledge of this feature. Please don't tell me to use BI's deepstack. I might one day, but for now, I want to use IVS as my main trigger for BI (Onvif trigger). Also, if you're going to yell at me because what I'm asking is "in the manual" or "in a thread" somewhere, then just tell me where. I've spent several days, many hours, searching, looking at IVS related videos, downloading a bunch of documents related to Dahua IVS and still have questions that I hope experienced folks can answer. Much of the info I've seen is not clear all the settings and many assumptions are made that you already understand a particular function/feature.

In the attachment below I've set up 2 tripwires. 1 is the zigzag line that should catch a car or person on the sidewalk either direction. The other one is supposed to catch (human) walking in my yard. The street/sidewalk works "ok" but occasionally there is a person walking on the other side of the street that it misses. Is this the best setup for what I want to catch or what would you do? I just want to catch cars on the street and people walking on the street or sidewalk.

In the IVS settings, does the tracking duration mean how long the camera will track something it's detected before it stops? My Dahua stops tracking after about 10 seconds or so.
I understand the tracking size can be adjusted to ignore small things (dog/cat/squirrel) and made large enough to trigger something the size of a human. There is another setting called "tracking target size ratio". How does that differ or compliment each other? It seems like those two settings are both similar to me.

When a car passes on the street when the car comes from left to right the camera does a decent job of tracking, but occasionally goes to far to the left. When a car comes from right to left, it does a poor job of 'catching' the car before it goes past the fence shown in the snapshot. The cars aren't going very fast because that direction is the end of the cul-de-sac. How can I improve how quickly the camera triggers, from left to right in this case, to catch a car sooner?

Under the Record/Record Control menu there is an option to pre-record, which I have set to 5 seconds. Shouldn't that setting capture the car passing in the scenario above?
Also under that section is "Max Duration", which I've set to 15 minutes. Does that mean as long as there is motion, it should record up to 15 mins? It doesn't.

Thanks for any help. I have 3 Dahua cams, my newest is the SD4A425DB-HNY which I'm working with now. I communicated with Andy @ EmpireTech and he did send a video tutorial but it was more basic than the questions I still have. I've found various documentation from other Dahua cams but not any single comprehensive document.

 

[wittaj]

Nobody is going to tell you to use Deepstack as that won't help with a tracking camera as all of the tracking is done within the camera....

First off, and I am sure you are aware from your other posts, make sure you are using Internet Explorer. Any other browser and the track time will default to 15 seconds.

Without seeing a video clip we have no idea if your tracking is only lasting 10 seconds because of an obstacle in the way or it moves out of the view too quickly, you are using a different browser and stops after 10-15 seconds, or if it is settings of the camera.

Ideally you need an SD card in the camera where you can turn on the IVS rules and see why it loses track (and I know from your other posts that you do albeit with issues).

You are incorrect on tracking size - the tracking size is how large do you want the object to be when it is tracking - it is not the size to not track dogs and animals - that is part of the object size and AI...

So for example if you have a tracking size of 50, that means the camera will zoom in to try to make the object 50% of the total field of view while it is tracking.

So if the tracking size is too small, the person will stay small in the field of view as it tracks. Make it larger and you get it to zoom in. But too high of a tracking size and it is easier for it to lose tracking. And too large and you get butts and boobs and no faces in the field of view LOL.

So you have to play with it so that as it zooms in and tracks, it reliably gives you what the goal is for that camera. Tracking size may be 30 or 50 or 70 or any other number. Most are around 50 or so.

Use BI for your recording and the SD card as a backup. The pre-record is how long it records (I know it is poor English translation LOL). The max duration is how large of a video file will it create, not how long it will record for. So if it was recording for 30 minutes, it would break it up into two 15-minute clips.

That is kind of a difficult field of view to reliably track cars. By the time the car gets in the frame, it is too much of the side of the car and it will have trouble tracking. It needs to be a further down the road field of view to capture more the front of the vehicle than the side.

Based on your IVS lines, it will have trouble catching on the other side. An object has to have half of its body cross an IVS line in order for it to trigger reliably. You need to make them further up into the yard on the other side of the street.

With that said, you are also trying to do too much with one field of view. It is why we use fixed cams as spotter cams to direct the camera where to start looking. If you want to capture people on both sides, ideally you have a preset for your side of the road and a preset for the other side of the rode and have a fixed cam direct the PTZ to the correct preset.

Tracking works best when the object is in the center of the frame, not the edges.

Also it is a matter of getting the brightness/contrast and target ratio settings correct.

You need to adjust gamma and the others as well.

Ideally for an intrusion box or tripwires, you should have the initial field of view be such that the camera doesn't have to initially pan too much up/down or left/right to get the object in the center of the screen to start tracking. The closer the object is to the center of the image, the better the chance that it will track correctly. An entire Field of View intrusion box can cause it to latch on to the wrong item, as can a tripwire to far on the left or right. This is probably what you are experiencing.

I always knew that you shouldn't chase a bright picture - it looks nice and people migrate towards a brighter TV for example, but upon closer examination, most images need to be toned down in order to get all the details. You will be surprised how much changing a parameter like gamma could impact tracking. For example, if you have a pesky tree or something in the middle of the view during an autotrack, just by changing some image parameters you can get autotrack to pass it. Making the image a little darker at night actually helped with tracking someone across the street, which was opposite of what I thought you would think to do. So add some contrast to your image and see if it improves.

I have a yard lamp post that more times than not autotrack would get stuck on it as someone was walking and the autotrack would only go so far. Because my image has soo much contrast (bright white concrete a third, blacktop road a third, grass a third), knocking down the gamma made the lamp post not be so "trackable" lol, and along with that I turned of PFA and that gave it just enough time to retrack the person walking past the lamp post. The camera may still autotrack the lamp post when a small kid goes by, but an adult it is autotracking past the lamp post. Most see better results if the contrast number is 8-10 higher than the brightness number.

Sometimes the initial track will wonder off to nowhere. The reason it starts looking upward or left or right is usually because the intrusion box is too big or the tripwire isn't close to the center so the camera identifies the object before it is in the center of the field of view and then sometimes something else matches the "algorithm signature" of the initial object and then starts trying to track something that isn't there. Adjusting the field of view and the locations of the IVS rules to be closer to the center can fix that.

Autotracking PTZs are great, but they have limitations like everything else. Installed in a wrong location or using default/auto settings or with fields of view that do not give it a chance will be problematic.

 

[marklyn]

Nobody is going to tell you to use Deepstack as that won't help with a tracking camera as all of the tracking is done within the camera....

First off, and I am sure you are aware from your other posts, make sure you are using Internet Explorer. Any other browser and the track time will default to 15 seconds.

Without seeing a video clip we have no idea if your tracking is only lasting 10 seconds because of an obstacle in the way or it moves out of the view too quickly, you are using a different browser and stops after 10-15 seconds, or if it is settings of the camera.

Ideally you need an SD card in the camera where you can turn on the IVS rules and see why it loses track (and I know from your other posts that you do albeit with issues).

You are incorrect on tracking size - the tracking size is how large do you want the object to be when it is tracking - it is not the size to not track dogs and animals - that is part of the object size and AI...

So for example if you have a tracking size of 50, that means the camera will zoom in to try to make the object 50% of the total field of view while it is tracking.

So if the tracking size is too small, the person will stay small in the field of view as it tracks. Make it larger and you get it to zoom in. But too high of a tracking size and it is easier for it to lose tracking. And too large and you get butts and boobs and no faces in the field of view LOL.

So you have to play with it so that as it zooms in and tracks, it reliably gives you what the goal is for that camera. Tracking size may be 30 or 50 or 70 or any other number. Most are around 50 or so.

Use BI for your recording and the SD card as a backup. The pre-record is how long it records (I know it is poor English translation LOL). The max duration is how large of a video file will it create, not how long it will record for. So if it was recording for 30 minutes, it would break it up into two 15-minute clips.

That is kind of a difficult field of view to reliably track cars. By the time the car gets in the frame, it is too much of the side of the car and it will have trouble tracking. It needs to be a further down the road field of view to capture more the front of the vehicle than the side.

Based on your IVS lines, it will have trouble catching on the other side. An object has to have have of its body cross an IVS line in order for it to trigger reliably. You need to make them further up into the yard on the other side of the street.

With that said, you are also trying to do too much with one field of view. It is why we use fixed cams as spotter cams to direct the camera where to start looking. If you want to capture people on both sides, ideally you have a preset for your side of the road and a preset for the other side of the rode and have a fixed cam direct the PTZ to the correct preset.

Tracking works best when the object is in the center of the frame, not the edges.

Also it is a matter of getting the brightness/contrast and target ratio settings correct.

You need to adjust gamma and the others as well.

Ideally for an intrusion box or tripwires, you should have the initial field of view be such that the camera doesn't have to initially pan too much up/down or left/right to get the object in the center of the screen to start tracking. The closer the object is to the center of the image, the better the chance that it will track correctly. An entire Field of View intrusion box can cause it to latch on to the wrong item, as can a tripwire to far on the left or right. This is probably what you are experiencing.

I always knew that you shouldn't chase a bright picture - it looks nice and people migrate towards a brighter TV for example, but upon closer examination, most images need to be toned down in order to get all the details. You will be surprised how much changing a parameter like gamma could impact tracking. For example, if you have a pesky tree or something in the middle of the view during an autotrack, just by changing some image parameters you can get autotrack to pass it. Making the image a little darker at night actually helped with tracking someone across the street, which was opposite of what I thought you would think to do. So add some contrast to your image and see if it improves.

I have a yard lamp post that more times than not autotrack would get stuck on it as someone was walking and the autotrack would only go so far. Because my image has soo much contrast (bright white concrete a third, blacktop road a third, grass a third), knocking down the gamma made the lamp post not be so "trackable" lol, and along with that I turned of PFA and that gave it just enough time to retrack the person walking past the lamp post. The camera may still autotrack the lamp post when a small kid goes by, but an adult it is autotracking past the lamp post. Most see better results if the contrast number is 8-10 higher than the brightness number.

Sometimes the initial track will wonder off to nowhere. The reason it starts looking upward or left or right is usually because the intrusion box is too big or the tripwire isn't close to the center so the camera identifies the object before it is in the center of the field of view and then sometimes something else matches the "algorithm signature" of the initial object and then starts trying to track something that isn't there. Adjusting the field of view and the locations of the IVS rules to be closer to the center can fix that.

Autotracking PTZs are great, but they have limitations like everything else. Installed in a wrong location or using default/auto settings or with fields of view that do not give it a chance will be problematic.

Wittaj, you are the best. Thank you for such a detailed explanation, enough that I'll have to read and re-read it, but it's excellent. I've definitely learned something and will make suggested changes. I may ask some more questions as I come across new features or issues. Thanks again!

 

[marklyn]

Ok, while I'm still futzing with everything wittaj gave me above, I got distracted and cannot find the culprit to a new issue.
I have my IVS (tripwire and intrusion) setup to trigger for cars/people walking/driving in front of my house. That seems to be working but I've got a lot of "general" recording going on. I don't have motion detection (video detection) turned on so I don't know why I have so much constant green showing up in my playback bar. Any ideas? Oddly enough when there is an event listed in playback it's usually in the general playback right before the trigger for the IVS where the camera captured something. For example, if a car is driving by, the event almost always misses it, but if I rewind a bit further in the green (General) area, it's there. I assumed if I turned off the motion detection under the Event menu, I would only get recordings for the IVS stuff I have turned on.

 

[wittaj]

Many of us have seen that even though it isn't supposed to, BI will turn on MD and SMD on the PTZs.

The "hack" we have found is go into MD and SMD and make it the least sensitive and in the area do just one tiny block and set a schedule of the min it will allow so that way if it is sending back a trigger code it shouldn't do it then.

 

[bigredfish]

Ok, while I'm still futzing with everything wittaj gave me above, I got distracted and cannot find the culprit to a new issue.
I have my IVS (tripwire and intrusion) setup to trigger for cars/people walking/driving in front of my house. That seems to be working but I've got a lot of "general" recording going on. I don't have motion detection (video detection) turned on so I don't know why I have so much constant green showing up in my playback bar. Any ideas? Oddly enough when there is an event listed in playback it's usually in the general playback right before the trigger for the IVS where the camera captured something. For example, if a car is driving by, the event almost always misses it, but if I rewind a bit further in the green (General) area, it's there. I assumed if I turned off the motion detection under the Event menu, I would only get recordings for the IVS stuff I have turned on.

It looks like you have General (full time) recording set.

 

[marklyn]

It looks like you have General (full time) recording set.
View attachment 170199 View attachment 170200

Oh. I turned it off on the time plan but didn't turn it off under the storage because I assumed if it wasn't in the time plan it didn't matter.
I'll try turning both off.
Thanks!!!