Computer vision for driver assistance systems

Abstract

The objective of the proposed thesis is to illustrate the training, validation and evaluation of vehicle detection algorithms using computer vision and deep learning methods, and vehicle tracking in video sequences. Object detection poses a major challenge in localizing the object and finding its exact coordinates, in addition to determining its presence in the image. Objects in varying scales must also be detected in the image, and the algorithm chosen for object detection must be feasible for real-time detection. The research work focuses on a traditional machine learning approach and a deep learning approach for overcoming the object detection challenges, and object tracking algorithms, to improve the accuracy of object detection.