utkuucak/utku-sinan-undergrad-thesis

Implementation of a self driving car using computer vision and machine learning techniques.

Utku-Sinan-Undergrad-Thesis

Implementation of a self driving car using computer vision and machine learning techniques.

OBSOLETE - /main : main project code, .py modules are stored here - Since we are going to split the code, this directory is not very suitable

/pc : Code that runs on pc

/pi : Code that runs on Raspberry Pi

/vids : video files are stored here

/train: training data are stored here

utku-sinan-undergrad-thesis's Issues

Calculate the angle between middle_line and position_line

A function is necessary to find out whether the car is on the left side of the middle line or the right side. This function is necessary for the controller implementation.

Warping(Could be a solution)

Warping can be considered in the future

Rasp Road seg., The road must be perfect otherwise runtime err

As mentioned #7 the program gives runtime error if it doesnt recognize a line at any time. This is a challenging issue for real time road segmentation. The road papers should be perfect or functions needs exception handling.

Improve directory structure

A conventional project directory structure is better for long term project management. It would be beneficial to reorganize project files according to Pyhton project best practices. This is NOT an urgent issue and I will look into it occasionally.

Implement draw_position_line

Write a function to print a 'position line' that would being from the middle point of the south border of the image and extend upwards. It will be used to calculate the angle between middle lane and decide whether car needs to steer left or right.

Create Server - Client Pairs

Two server - client pairs are required for communication between laptop and Raspberry Pi.

Server at Pi would send frames captured from camera module to laptop in real time and client in laptop would receive them and pass it to image processing script.

Server at laptop would send the steering information to Pi client and control the car's motors accordingly.

Extrapolation and Averaging with lines with gaps

Extrapoalation and averaging operation works fine with normal lines but with the lines with gaps it sometimes gives weird outputs

NoIR Camera power

Camera draws too much power, the battery must be replaced w 2A

Region of interest for different videos

RoI was defined specificly to the first video. However for different videos their RoI is actually a litle bit different. This case must be checked.

Video bittiğinde pencere donuyor

Lİnux'da bu şekilde en azından. Heralde kodun altına bir pencere kapat komutu gelmesi gerek ama issues kısmını da test etmek için bu issue yu açıyorum.