Driverless Cars-why the World isn’t ready
Lane detection algorithm
In this technological era as many AI experts and Tech companies have continuously made great strides towards the improvement of intelligent transport systems using AI algorithms to achieve accurate lane mark and object detection. Is the World really ready for a fully autonomous transport system?
I will dive directly into a simple lane detection algorithm using OpenCV. The sample code source can be found at: https://github.com/Jonathan-WS/self_driving_cars
In this simple lane detection algorithm, we can find functions allowing us to do the following: Grayscale conversion and canny edge detection of an input image, cropping to a region of interest in an image, the detection of the edges in the cropped input image and the use of Hough transforms to detect marked lines in the input image.
While I won't go into the detailed mathematics of the above stated functions, for which some are predefined Python Standard Libraries. For example, the concept of a canny edge detection algorithm which detects how strong the gradient variation in the image’s color function is. In the Canny function we also use parameters such as the threshold which indicate generally how strong an edge of a lane mark must be in order to be detected in terms of its intensity level.
So applying the lane detection algorithm on the selected input image gave the following output:
AI could be the silver bullet to some cons of autonomous vehicles. Further more, Autonomous vehicles require redundant camera, radar(radar for light waves) and LIDAR, in order to mimic the ability to see the marked lanes and the surrounding objects. The sensors offset one another’s strengths and weaknesses. While Cameras are good at object detection they struggle with low light. LIDAR works perfectly in every lighting condition but suffers in scenarios like bad weather (snow etc).
References:
https://www.youtube.com/watch?v=4zHbI-fFIlI
https://www.youtube.com/watch?
https://docs.opencv.org/trunk/da/d22/tutorial_py_canny.html
https://alyssaq.github.io/2014/understanding-hough-transform/
