Obstacle Identification Through Fast Vector Analysis

During road travel, obstacles can impede productivity or durability for many different vehicles and render discomfort or injuries for the people within. Using remote sensing techniques, information from the surroundings can be acquired and analysed to identify obstacles ahead. The subsequent analysis can create a decision support for how the vehicle or driver should act upon encountered obstacles, through either autonomous control, guidance to the driver or a combination of both. In this paper, an experimental setup was created to mimic an obstacle in the shape of a speed bump on a flat road. An RGB-D camera was used to acquire information while travelling towards the speed bump. Afterwards, the acquired information was analysed by an estimation of the normal vector for each point in a 2D depth map. The resulting data from the experiments had sufficient resolution, speed and quality to retrieve proper identify obstacles or targets indoors with an accuracy of 2%. Obstacles were measured and identified in less than 20 ms where processing time mainly comprised data transfer from the USB-bus. The obstacle identification can be used to e.g. actively control the vehicle suspension, send feedback to the driver about obstacles ahead or optimise speed and direction for autonomous vehicles.