Object Detection and Segmentation for Scene Understanding via Random Forest
Bisma Riaz Chughtai, Ahmad Jalal
Abstract
In recent days, object detection become a vast topic in computer vision. Accurate object detection and scene understanding is not an easy task due to illumination, viewpoints, and color intensities. Visual features like color, texture, boundaries, and shape, make an image different from another image. The main goal of scene understanding is to machine work like a human and understand the visual information of an image. Currently, researchers working on novel approaches in this field to make a better understanding of the scene. Computer vision portrays a major role in different applications such as health, safety, security surveillance, traffic monitoring, autonomous driving car, object recognition, and tracking. In this paper, we work on a meaningful understanding of an image in the scene. To understand the scene we have done region-based segmentation, and for object detection and labeling, we use the tensor flow algorithm, for geometric features mean of each pixel, harry corner edge detection, and scale-invariant feature transform descriptor. And then object recognition by using random forest. We have performed this experiment on UIUC Sports dataset. The presented model achieved 89.45% recognition accuracy.