Shadow based Time Prediction in Video Sequences using Hough Transform

Abstract

Identifying informative objects under the presence of shadows is a difficult step in computer vision applications. While capturing images and videos cameras plays a vital role in gathering information in the absence of paid persons. But in low light conditions camera can result in poor quality videos. And also low light can lead to camera shake and blurred images and videos. Forensic Identification Services (FIS) face difficulty in case of emergency conditions for finding the time of the incidents.In this paper, we give more importance to shadow part of an image/video. Initially detects the part of the image that gives object and the related shadow by using background subtraction method. To detect the shadow part YCbCr color spaces are employed. Morphological operations are used to extract the detected shadow regions and remove the unwanted tine objects. Using Hough Transformation line drawn over the object and the associated shadow to extract the length, angle between the object and the shadow. Time estimation is done based on the features like length and angle.