| The Intelligent Vehicle (IV) is being established as a major component of the Intelligent Transportation Systems (ITS). The intent of the IV is to improve significantly the safety and efficiency of motor vehicle operations by reducing the probability of motor vehicle crashes. In this thesis, a proactive exterior airbag system is proposed to address the issue of limiting the severity of a crash when it becomes unavoidable.; Situational awareness, or knowing where other objects are, is an important capability for IV. If the relative position and speed of other objects is known, a course of action can be planned that might involve warning drivers, recommending control actions, intervening with driver control, introducing automated control of the vehicle, deploying the proactive exterior airbag system in hazardous situations.; Accurate object detection, recognition and tracking system is the essential part of situational awareness. In this thesis, we present a vision-based approach to detect and recognize various objects, calculate their motion parameters, and track multiple objects by using a sequence of color images taken from a moving vehicle. Three kinds of object are investigated, including vehicles, traffic signs, and lane markers. A complete suite of algorithms for the detection of these most common objects in the driving environment has been developed. In the analysis of single image frame, the system detects potential objects by using their shape features and recognizes the objects by using neural network. Once the objects are recognized, they are tracked in the consecutive image frames. Tracking is done by combining simple image processing techniques with Kalman filter. To improve the computational speed, the analysis of the single image frame is performed every 10th full-size images, the rest of images will only be processed by the region of interest. Experimental results demonstrated a robust and accurate system in real time object detection, object recognition, object tracking, and object motion analysis over thousands of image frames. |
