artusindus/picar-x-line-following-with-opencv-and-picamera2

This project is a line following program for the PiCar-X, a small robot car kit built around a Raspberry Pi. It uses a camera along with the OpenCV image processing library and the Picamera2 camera interface to spot a white track on the ground in real time and steer the car so it stays centered on that path. The way it works is straightforward. The robot takes a picture from its camera, converts it to grayscale, and then applies a threshold so that the white track stands out clearly from everything else in the image. From that simplified black and white image, it calculates something called image moments, which is really just a way of finding the middle point of the white area. By comparing that middle point to the center of the whole image, the program works out how far off course the car is, turns that difference into a steering angle, and keeps the car driving forward while constantly correcting its direction. To run this yourself, you need the actual PiCar-X hardware, a Raspberry Pi, a Raspberry Pi camera module, and a power supply. On the software side you need Python 3 along with the OpenCV, NumPy, Picamera2, and PiCar-X libraries installed on the Raspberry Pi. Once everything is set up, you install the Python dependencies with pip, then run a single script to start the line following behavior, and you can quit at any time by pressing the Q key. Several settings in the code can be tuned directly, including the driving speed, how far the camera tilts, the brightness threshold used to detect the white line, the maximum steering angle allowed, and how sharply the car reacts to being off center. The project also shows a live view of the camera feed and the processed black and white image, which is useful for checking that the track detection is working correctly, and it includes handling so the robot shuts down safely. The author lists several planned improvements that have not been built yet, such as adjusting the threshold automatically based on lighting, focusing processing on a smaller region of the image, adding a proper PID steering controller for smoother corrections, detecting curves better, and filtering out noise. The project is released under the MIT license.