based on Image Processing using Convolutional Neural Networks + OpenCV.
- The project aims to build a monocular vision autonomous car prototype using Raspberry Pi as a processing chip interfaced with a camera. -It is based on Image Processing using Deep learning + OpenCV,which classifies between the various images.
- Raspberry Pi CAM was used for all the real time image input .
- Images were simplified using OpenCV - hough transform,canny edge detection to make it easier for our neural network to identify features and colour/region masking to avoid unnecessary striking features to interfere in the model
- Raspberry Pi 3 was used for nominal real time image processing using open CV which is powered by the 12V adapter to avoid input power fluctuations.
- The model was trained using KERAS library with 2000 images each for "left, right, straight, sharp left, sharp right,stop" to follow a track.Sample images of the above categories in provided in the sample data section of the repository.
- However the Arena for which the model trained to run for has limited features like stop sign ,no entry,steep left/right,front and stop sign.
- In the cases where deadlines that are marked by black lines on the arena ,are to be handled IR sensors are arranged in front of the bot to detect them.
- The motors were controlled using a microcontroller, Arduino UNO which is serial communicated with raspberry pi.
- This robot could also be used as a remote control car to race around the track,when we manually input the keys,using serial communication of arduino and raspberry pi,And can also be used as a line follower bot using IR sensors and arduino.
