Estimating the distance to an object based on image processing

When it comes to vehicles and traffic, the first priority is the safety of the driver and that of pedestrians. With the evolution of technology, the speed of the cars and the traffic increased. This means that there is a need to predict dangers which can arise while someone is driving. Another direction is to develop autonomous cars that can face the present and the future traffic conditions in order to increase the safety and the fluency on the roads. So, in the present paper it is presented a method of determining the distance to a car that drives in front of our own car. The relative speed is also possible to be determined with respect to it. The image, taken with a front camera, is classified using a Convolutional Neural Network, namely YOLO [1] (You Only Look Once) and after the searched object is detected, the distance is estimated counting the number of pixels in a bounding box which fits the detected object. The distance is further corrected by using Canny edge detection and HSV color space. Experimental data are presented in the paper and the results are commented for conclusions.