Parking mode

your camera (in "motion detection" enabled mode) will not capture the first part of event. you will not see the approaching car, only the recessive car.
 
jokiin said:
what is the environment like though, motion detect on a home CCTV system is nothing like a car system, your home cameras are in a fixed location and only look at a specific scene, you can calibrate a CCTV system to ignore what you want and concentrate on what you want, the system in your car needs to work in any location, the view is ever changing, it won't ever work like your home CCTV system does
Click to expand...

Absolutely spot on but, in any MD system I have seen, there is the option to decrease/increase Sensitivity....
That's all I want......I just want to tune the MD to ignore flies at 100 yards, wind blowing bushes,trees etc and record peeps and vehicles going past the camera.
I am I expecting too much?
 
Head Gardener said:
I am I expecting too much?
Click to expand...

basically yes

remember this is motion detect, not parking mode, if you are wanting to have the motion detect on all the time, tuning it to suit when parked would make it useless when driving, and vise versa, I take it you want to manually toggle to motion detect on everytime you park the car???

if the motion can be automated to only function when parked then sensitivity adjustments would be more beneficial, at the moment there are people that use it all the time, your adjustment would break it for them so it would become something that needs to be turned off/on all the time
 
Jokiin, now we are getting somewhere....I see that 'parking mode' records the event after triggering and not pre event.
However, I assumed that MD mode was the same mode as used in the Aston Martin vid I posted earlier in this thread....
I just thought the A119 would give me the same facility?
I am more than happy with the A119 in every aspect apart from the MD mode....
All I would wish for is an option to adjust the sensitivity of the MD mode, at present, my camera may as well not have MD, and just be set to continuous recording, what ever the res/FPS.... All that some control of sensitivity would allow me to do would be to go straight to the event when reviewing the recording on my PC...should an 'incident' occur, rather than FFW thru 10 hrs footage...
If however, there is no hope of a firmware update to address this, I still think the A119 is a great piece of kit!
 
Head Gardener said:
If however, there is no hope of a firmware update to address this, I still think the A119 is a great piece of kit!
Click to expand...

It can be mounted on a network of sensors PIR inside the car, to powered dash cam every move, 1 min(more or less).In this way the movement is more effective referral
 
doru3160 said:
It can be mounted on a network of sensors PIR inside the car, to powered dash cam every move, 1 min(more or less).In this way the movement is more effective referral
Click to expand...

More info please Doru!
 
there is only one rule in this world: you change it or adapt to it. :)
 
Head Gardener said:
Jokiin, now we are getting somewhere....I see that 'parking mode' records the event after triggering and not pre event.
However, I assumed that MD mode was the same mode as used in the Aston Martin vid I posted earlier in this thread....
I just thought the A119 would give me the same facility?
I am more than happy with the A119 in every aspect apart from the MD mode....
All I would wish for is an option to adjust the sensitivity of the MD mode, at present, my camera may as well not have MD, and just be set to continuous recording, what ever the res/FPS.... All that some control of sensitivity would allow me to do would be to go straight to the event when reviewing the recording on my PC...should an 'incident' occur, rather than FFW thru 10 hrs footage...
If however, there is no hope of a firmware update to address this, I still think the A119 is a great piece of kit!
Click to expand...
Of course, motion detection can be adjusted, we have done this in A118C.

Regarding the motion detection mode, one thing we' d like to remind, while in motion detection mode, do not stop recording manually, otherwise it will not detect motion anymore(not start recording even there is a new motion).
 
  • Like
Reactions: Mtz
A while ago I wrote motion detection script for IP Cameras... I use it for my cameras around the house.

I can tell you getting it right it is not easy. My motion detection consumes about 5%-10% of CPU on a modern x86 based architecture. That is probably 2-5 times more than what is available on the dash cam.

In reality I believe my script is doing much better (for motion detection CPU usage) than say zone alarm or similar product.

To start with, it is not possible to have it perfect in all conditions (day or night) without having a cluster of servers processing the video data. Another problem is pre-buffering video, that takes RAM and CPU cycles, as well as none of the SoC actually encode the video on CPU, so the hardware h264 encoder DSP must have that facility (hence probably why Novatek does not do it, but Ambarella does). Motion detection itself probably comes from encoder DSP and not CPU, thus making it more rigid.

If you are worried about events before the incident, best bet is to invest in 128GB card (about 15hours of video), or another camera, have two cameras running in following configurations: 1 running in full res @ 1440p, another running @ 720p/15fps (or whatever lowest settings is). Have the 1440p one is running in motion detection, while the 720p camera running constantly.

For those interested in my implementation of motion detection/recording script for IP Cameras, here is the evolution:
My first attempt was to use a perl script that was calling imagemagick and comparing frames. That was very in-efficient and inaccurate (consuming ~50% CPU with 320x240 image on Core2Duo).
My second iteration was using OpenCV with python and getting a running average frame and subtracting current frame from running average image, then drawing contours around the result and judging form the size and position of contours the degree of motion. That was twice more efficient and had better accuracy. I could also bolt on the area of interest (a polygon), checking if the motion happens in the area of the interest.
My third iteration was using OpenCV with python, but instead of running average and subtraction I used BackgroundSubtractorMOG2 which turned to be a very efficient and noise resistant motion detection algorithm. In same manner as running average method I could utilise the area of interest polygon and remove those pesky trees from being detected.This allowed me to use only 5-10% of the CPU on something like low end Pentium (LGA1150) while using 720p RTSP sub-stream as source.

How do I record when motion happens? I simply launch ffmpeg instance, which converts RTSP HD stream into mkv (without transcoding). The ffmpeg only takes about 1-2% of CPU. There is no prebuffering happening (it takes 100% CPU to "prebuffer" 1080p with via OpenCV on Core type of CPU). I have an idea how to do it efficiently. Dumping RTSP stream into mkv without transcoding is very efficient, I could run two captures for prebuffer time period side by side, but offset half way from each other, and then choose and save the relevant one (longest running one relative to moment of motion detection) as separate file.

TL,DR; motion detection is hard.
 
Head Gardener said:
More info please Doru!
Click to expand...
If indeed it manages to implement motion detection function, it will have only drawback to detect movement in the visual field of dash cam.Whith a group of PIR sensors can be detected 360grade move around in the car.If person scratch up the car from the back, then motion detection not help anything.Dash cam is new, I have not had time to experiment, but I will come back with concrete results as you will be able to realize practical idea,
as I am interested in parking mod.I still managed to manufacture a filter CPL.No profi show, but it does the job
For now I enjoy your purchase




 
Last edited:
Head Gardener said:
And during a 10pm till 08am test using MD I had 10 hrs solid, minute after minute, collection of 3 minute files to look thru....(32 Gbs worth)
Click to expand...
If your car's security system has a bright flashing LED on the dash, it may be constantly triggering the motion detection at night?

My SGZC12RC saves clips starting 5 seconds before the event that triggered the motion detection, e.g. a passing bird or car, etc. If nothing else happens after the car passes from view, the saved clip is roughly 45 seconds in duration. Then a while may pass before the next event triggers the next saved clip.

image.jpeg

If motion is near continuous, e.g. tree branches moving on a windy day, I end up with a succession of 2 minute clips.

If you want pre-buffered motion detection, the cheapest route is the MG600, the generic relative of the SGZC12RC, not as good as the SG but good value at £74 ...

https://www.ebay.co.uk/itm/282124155147
 
You must log in or register to reply here.
Back
Top