Simulation of an Autonomous Vehicle Control System Based on Image Processing

Control of autonomous vehicle is a developing subject in the computer vision community. Control of autonomous vehicles has many applications in city cars, transportation trucks and specially in agronomy industry. We present preliminary results of a proposed control system simulation for an autonomous terrestrial vehicle. Our goal is to use an autonomous vehicle for inspection of plantations in agriculture. We constructed a prototype vehicle with a 3D printer and two motors. This prototype vehicle was used to obtain a dynamic model and, with this model, we developed a controller for navigating the vehicle. The considered visual features are colors of images, specifically, we are interested in detecting road-color. We have constructed a detector based on the color histogram to determine whether a pixel belongs to the road-color class or not. Depending on the number of pixels that belongs to this class, in a region of interest, the controller takes action through the DC-motor of the vehicle. For this simulation, we used a synthetic video database, where objects move toward or away from the camera. We have tested our proposal with video sequences using the model jointly with the controller and we have demonstrated that our proposal can avoid obstacles that move on a straight line or located in random positions on the road. We have compared our proposal with other method based on gradient. Our proposal can perform better than the method based on gradient in this specific task and in the considered database.