Regresar

Vehicular Obstruction Detection in the Zebra Lane Using Computer Vision

Abstract:

Computer Vision and Image Analysis are used in researches with an objective of extracting information from a set of scenarios. Multiple researches with varying objectives like vehicle speed detection, traffic density estimation, vehicle counting, or in general, observation of behaviors of multiple objects, have been applying Computer Vision. This research paper is about utilizing Computer Vision for obstruction detection by observing temporal state of vehicles situated in a pedestrian crossing lane. The researchers gathered data by taking videos of real traffic in a road containing a pedestrian crossing lane (PCL). The method starts with a pre-processing phase wherein the image was de-noised, converted to grayscale and derived the Image Binarization, and establishment of the PCL for region of interest (ROI). Connected components are extracted then assigned its own structure with corresponding properties called 'track'. Tracks are monitored by using Kalman Filter and Hungarian Algorithm. Then, Ray-Casting algorithm is applied to determine if an object violates the traffic rule. For violators, a snapshot will be taken and determine the license plate. Based on the result, True Positive Rate of 65.28% and True Negative Rate of 98.26% were obtained.