Object Detection

For object detection we used a combination of the front distance sensor and classical computer vision methods.

If the distance sensor detects an obstacle at less than x cm, then the decision center of the car stops it, independently of what the object may be.

To classify the object, we use a combination color segmentation/thresholding and morphology operations. If we detect a pink blob of a certain ratio / dimensions then we classify it as a pedestrian. If we detect yellow and red blobs then it is a road barrier. In any other case, we assume that it was a car (as the competition track is a controlled environment it is acceptable to make that assumption).