Congrats. I’ve been lurking since I have nothing substantive to add. But agree completely with @cyclicalobsessive and @jimrh - get the minimum expected done. If you have time after that you can add additional bells and whistles.
Looks like you’re a long way to an MVP (minimal viable product). Great work!!!
/K
Perhaps this is over thinking it?
If you gave me the
I expect
and all those other lines you are computing may only be good for proving to a human you can process an image for human consumption. They may be of redundant value or even a drag on a basic control algorithm. Later with more processing resource, perhaps more points improve confidence or allow for broken middle lane lines, or the missing lane lines when passing an exit.
For initial proof of concept, without a GPU to do all the calculations, and running an interpreted/bytecode language such as Python - start simple and optimize later.
As another investigation, running a Kalman Estimation Filter on p1, and p2 each detection loop, might cover the temporary loss of both lane lines condition.
Can you elaborate what you mean with dH? 
I have two generell questions about a raspberry pi that I just can’t figure out.
First:
To change the resolution:
sudo raspi-config
(I am not near my system and I am doing this from memory. Names may be slightly different.)
Reboot and see what you get. Retry as needed to get it right.
I noticed that to get a reasonable resolution, I had to set up a desktop, and set that resolution to something reasonable. Since VNC simply “mirrors” the desktop, you need a configured desktop to mirror.
Second:
What kind of a computer are you using?
On a PC you can use FileZilla to set up a secure connection to the 'bot to transfer files.
Another thing you can do with VS Code is to set up remote development. that puts some software on the 'bot and allows you to run VS Code on your main system and develop code directly on the 'bot itself.
You will have to look that up yourself as I am not near my system right now.
So the one method to change the resolution is with peripherals and a monitor? But that also means that I need to have the monitor connected at all times if I want the resolution to not change back, right?
I am using a win10 pc. Nothing special. Okay FileZilla ist good to know. I am already using VS Code with the remote ssh extension to code on my pc and run the code on the gopigo
No. You don’t need the monitor all the time, just to configure the desktop resolution.
You can use VNC all other times.
As far as huge picture resolution is concerned, you might need to look at wherever the camera/video is configured to set that differently.
You do not need to attach a monitor/keyboard, just ssh in, and make changes:
Use scp secure copy command but FileZilla is easier
dH Delta horizontal = x1 - x2 (pixel)
dV Delta vertical = y1 - y2
Oh my god… It worked… Thank you so much!
Nice, Filezilla also works 
Thanks you both!
Since I cannot see what you are doing, I’m not sure what resolution you’re talking about.
If everything else is OK and its the CAMERA or VIDEO that’s taking too much room, you can change that by adjusting the /etc/uv4l/uv4l-raspicam.conf file.
Somewhere between lines 50 and 60, (depending on the version of raspicam and the config file you are using), you will see something that looks like the following:
(this is from the current version of raspicam’s uv4l-raspicam.conf file)
encoding = mjpeg
# width = 640
# height = 480
framerate = 30
#custom-sensor-config = 2
With the width and height lines commented out, the video image is HUGE - filling the entire screen/browser width.
Un-commenting these lines makes the image smaller, but still too big for what I want.
The original version of the config file as shipped with GoPiGo O/S 3.0.1 has this:
encoding = mjpeg
width = 320
height = 240
framerate = 15
#custom-sensor-config = 2
Which produces an image that’s a bit too small for my taste, and a bit blurry, so I increased the size by half, and upped my framerate to 30.
Viz.: (my current settings)
encoding = mjpeg
width = 480
height = 360
framerate = 30
#custom-sensor-config = 2
. . . and that produces an image that is large enough, but not TOO big.
Maybe this will help?
Follow-up note:
In order to maximize the field of view, make sure the settings maintain a 4:3 aspect ratio.
Quick update: I spoke to my professor about the minimum result, that I need to achieve for my thesis. He said that the minimum result is the perception of the environment, which means lane detection and obstacle detection of certain objects. I don’t need to achieve lane keeping, but it would be amazing if I did. So I try to get the lane detection with turns going. After that I will search for a obstacle detection algorithm for cars or stop signs or pedestrians. After that my focus is mainly on writing the thesis and which the time in mind, I can think of improvements of any kind.
Correct me if I am wrong, but (AFAIK) a “stop sign” isn’t an obstacle, right?
A person in the road, or another car, or a cow, or a fallen tree - these are obstacles, but a stop sign (or a traffic light) isn’t an obstacle.
Or, am I missing something here?
Correct a stop sign isn an obstacle. I was just brain storming while writing this 
. I need to prove that I can detect certain things basically. It doesn’t really matter what I am trying to detect. The easier to implement the better.
Do you have to use the camera for obstacle detection, or can you use another sensor (e.g. ultrasound)?
/K
Absolutely!
Forgive me for being a bit long-winded, but the topic of engineering/over-engineering reminded me of a story I read in a 1950’s/1960’s era trade magazine for television repairmen of that day.
Yes, I am only using the raspberry pi cam v2 for that.