Image processing based vehicle to vehicle distance estimation

It is important for vehicle drivers that their distance from the vehicle in front is safe to be considered. In this study, a method is presented to estimate the distance to the vehicle in front. For the implementation of the proposed method, two cameras mounted at the ends of the roof bar on a vehicle are used. The optical axes of the cameras are equally inclined towards each other and intersect with each other. In the study, an Artificial Neural Network (ANN) model was developed to estimate the distance of the vehicle in front. To create a data set, information obtained from the coordinates in the camera images of the center point of the plate of the vehicle in front is used. For this purpose, 414 pairs of image samples taken from different horizontal positions and distances of the vehicle in front were used. 352 of these image samples were used for the training data set and 62 for the test data set. Attempts were made to determine the input variables, the number of neurons and the number of iterations of the ANN model, which gave the best results. In the experiments, ANN model, which uses the eight parameters created from the horizontal and vertical coordinate information in the camera images of the target point as input variables, gave the best results. This result shows that the difference between horizontal and vertical positions in the camera images of the vehicle in the proposed method is determinant in the distance estimation. The limitation of the proposed method was that the detection of the target point was done manually. With the determined ANN model, the mean absolute percentage error value of the results obtained from the test data was determined as 1.6961%. Accordingly, the proposed methodology is effective for determining the distance between vehicles.