The comparison between two methods of object detection: Fast Yolo model and Delaunay Triangulation

Image segmentation, object detection and classification are three closely related tasks that can be greatly improved when they are jointly solved by feeding information from one task to another. Different methods have been proposed by the researchers, some of which have given good results and others fail in certain circumstances. In our paper, we compare two techniques for recognizing moving objects in a video scene. The first approach is based on deep learning. We implemented the Fast Yolo model to detect objects. The second approach is based on the segmentation of objects, we used the Delaunay Triangulation method to recover homogeneous regions. We have combined the features of the HOG, color histogram, and GLCM associated with each object. The classification phase is carried out by Alexnet for both approaches. The experiment was carried out on several video clips of highways and local roads with different traffic and lighting conditions.